Evidence on postpartum antidepressant effectiveness is limited. We sought to determine the association between different timings of antidepressant initiation within the 0–6 months postpartum and mental outcomes at 6- and 18-months following childbirth.
MethodsUsing data from the Norwegian Mother, Father and Child Cohort Study (MoBa) linked to a sub-sample of MoBa Genetics, we included 1146 women with psychiatric history before and/or during early pregnancy: 154 (13.4%) antidepressant initiators in the 0–6 months postpartum (37.0% within 0–3, and 63.0% within 4–6 months), and 992 (86.6%) with unmedicated new mental illness episode within same period. Depression/anxiety symptom outcomes at 6 and 18 months postpartum were self-reported. To control for measured confounding, we fit linear and modified Poisson models using overlap weighting, overall and stratified by polygenic score for antidepressant response.
ResultsAntidepressant initiators within 0–3 months postpartum had reduced severity of depression/anxiety symptoms at 6, but not at 18 months postpartum, compared with unmedicated women (weighted mean difference (MD): -0.23, 95% CI = -0.39, -0.06) and a 44% lower risk of having active, clinically relevant depressive/anxiety symptoms at 6 months (weighted RR= 0.56, 95% CI = 0.35, 0.91). No association was observed for initiation at 4–6 months postpartum at the 6-month assessment. The associations were evident only among women having high polygenic score for positive antidepressant response (weighted MD: -0.50, 95% CI: -0.84, -0.16), and that persisted at 18 months.
ConclusionsInitiation of antidepressant in first 3 months postpartum, but not in months 4–6, is associated with reduced severity of depression/anxiety symptoms at 6 months post-delivery. Effects were most evident among women with higher polygenic scores; however, these findings are exploratory and warrant larger studies.
Postpartum depression and anxiety are serious, often co-existing mental disorders experienced by 5–17% of women (herein defined as all persons who can physically give birth regardless of gender identity) in the first year after giving birth.1,2 Beyond being detrimental to maternal mental and somatic health, these disorders negatively impact mother-child bonding and offspring development,3 and they are major causes of maternal suicidal behaviour and self-harm, especially in the first year after their onset.4
Having a psychiatric disorder before and/or during pregnancy is major risk factor for postpartum depression and anxiety.5 Although depression and anxiety are prevalent in women before, during and after delivery,2 the first two months postpartum have been described as a period of increased risk for both new and recurrent episodes, and these were notably severe.6 The difference in clinical phenotype and symptom severity of depressive and anxiety according to their timing of onset postpartum,6,7 has impact on clinical decisions with regards to antidepressant treatment.
Antidepressants are generally recommended in treating postpartum depression and anxiety,8 albeit there is limited evidence on their effectiveness in clinical and real-world setting.9 The few available randomized controlled trials showed that antidepressant treatment in the first weeks postpartum may be effective in managing postnatal depression, especially in the acute phase; however the timing of randomization and treatment length varied across studies.9–13 Whether timing of antidepressant initiation postpartum poses different clinical benefit to maternal mental health outcomes, specifically in women with psychiatric history, has not been elucidated.
In a population of women with psychiatric history before and/or during early pregnancy, we aimed to evaluate the effectiveness of antidepressant initiation within the 0–3 and 4–6 months postpartum, as well as overall within 0–6 months, on a broad set of maternal mental health outcomes at 6- and 18-months following childbirth. Our secondary aim was to assess whether this association varied by clinical type, course and polygenic predisposition for maternal psychiatric history, and individual polygenic score for antidepressant response.
MethodsStudy setting and data collectionThis study is based on data from The Norwegian Mother, Father and Child Cohort Study (MoBa) linked to a sub-sample of MoBa Genetics, and the Medical Birth Registry of Norway (MBRN), via the unique personal identification number. MoBa is a nationwide, prospective population-based pregnancy cohort study conducted in 1999–2008.14,15 Postal invitations were sent to pregnant women all over Norway at gestation week 15–17, in connection to the publicly offered routine ultrasound. Approximately 41% of the invited pregnancies consented to participate.14,15 The cohort includes approximately 114,500 children, 95,200 mothers, and 75,200 fathers. Women were asked to complete several prenatal and postnatal questionnaires covering medical, socio-demographic, and psychological information about themselves and their offspring. This study is based on data from two prenatal questionnaires at gestational week 17 (Q1) and 30 (Q3), and two postnatal questionnaires completed at 6 (Q4) and 18 months (Q5) after birth. The questionnaires are available online.16 We used version 12 of the quality assured MoBa data files released for research. The project protocol has been preregistered (DOI 10.17605/OSF.IO/Q6VYH).
In MoBa, blood samples were obtained from both parents during pregnancy and from mothers and children (umbilical cord) at birth.17 Genetic data was retrieved from MoBa Genetics v1.0 for a subsample of women. Information on MoBa Genetics study cohorts, quality control and imputation is available online and in prior work.18,19 Summary statistics for depression, anxiety, and antidepressant response from recent genome-wide association (GWAS) studies in European populations were used for polygenic score (PGS) calculation. Details are given in Supplementary Material 1.
The MBRN is a national health registry containing information about all live births, stillbirths, and induced abortions after week 12. The registry comprises maternal medical records during prenatal care, as well as mother and child health at the delivery ward.
Study populationAs shown in Supplementary Figure S1, we included pregnancy-child dyads of women with a livebirth delivery who (i) returned two prenatal (Q1 and Q3) and one postnatal questionnaire (Q4), the latter within eight months post-delivery, and (ii) reported history of clinical depression, anxiety or eating disorders prior to and/or during early pregnancy. Eating disorders were identified using diagnostic algorithms based on the DSM-IV, as described in prior research,20 while depression and anxiety were self-reported, based on a list of long-term illnesses.16 In line with the criteria of the studies included in the Cochrane systematic review,9 having a new episode of depression and anxiety with onset postpartum was defined as: (i) having self-reported mental illness and/or antidepressant initiation in the 0–6 months postpartum, and (ii) having no clinical depression and/or other mental disorders and no use of antidepressants, antipsychotics or anxiolytics in the last two months before delivery.
OutcomesThe main outcome of self-reported depressive/anxiety symptom severity was measured at 6 and 18 months postpartum via the 8-item short versions of The Hopkins Symptom Checklist-25 (SCL-25) (SCL-8), a reliable screening instrument.21 Response categories (from 1="not bothered" to 4="extremely bothered", see Supplementary Table S1) were summed and subsequently divided by the total number of items, giving a final score from 1 to 4; lower score indicating lower symptom severity. To assess depressive and anxiety symptoms separately, we used scores from the SCL-8 subdomains of depressive (four items) and anxiety (four items) symptoms (Supplementary Table S1).22 In addition, we defined presence of active, clinically relevant symptoms of depression/anxiety (yes/no) using a stricter cutoff SCL-8 score ≥2.0 as well as a broader cutoff score (1.85) which has previously been validated among non-pregnant individuals.23
Our secondary outcome measures included anxiety symptoms, life satisfaction and self-esteem. Anxiety symptoms were additionally measured with items 3–5 of the Edinburgh Postnatal Depression Scale (EPDS) at 6 months postpartum. The EPDS is a widely used screening scale for perinatal depression,24 and anxiety.25 MoBa Q4 did not include all ten EPDS items. Each item response scored 0–3 on an ordinal scale, producing a total anxiety score of 0–9. Higher scores indicated worse symptomatology. We also defined presence of active anxiety symptoms by a score ≥6.0 (value of the upper quartile), as binary measure. See Supplementary Table S1 for details.
Life satisfaction was measured at 6 months postpartum via the 5-item Satisfaction with Life Scale (SWLS),26 and modelled as numeric score (range 5–35), where higher score indicates greater satisfaction with life. The 4-item Rosenberg-Self Esteem Scale (RSES) measured self-esteem at 6 and 18 months postpartum, with a total score of 4–16 (Supplementary Table S1). Lower score indicates lower self-esteem. We additionally defined a binary outcome measure for low vs normal-high self-esteem using the cutoff value 10 (lower quartile). Item response details for the SWLS and RSES and reliability assessment of all outcomes are given in Supplementary Material 1.
Antidepressant initiationExposure to antidepressants was defined as initiating a medication classified under the ATC group N06A, which was further divided into Selective Serotonin Reuptake Inhibitors (SSRIs) (ATC: N06AB), serotonin-norepinephrine reuptake inhibitors (SNRIs) (venlafaxine and duloxetine), and other antidepressants. The primary exposure windows were early (0–3 completed months) and mid (4–6 completed months) postpartum, and the overall period 0–6 months (Figure S1). These exposure groups were compared to women reporting, in MoBa Q4, new mental illness episode in the 0–6 months postpartum who were not medicated with antidepressants (hereafter, “unmedicated”). In MoBa Q4, women were specifically asked to report whether they experienced “mental health problems” up to 6 months postpartum post-delivery.
Measured confoundersConfounders were identified with the aid of directed acyclic graphs, and included pregnancy characteristics (i.e., body mass index (BMI) pre-pregnancy, parity and gestational weight gain); maternal sociodemographic and lifestyle factors (i.e., country of birth, age at delivery, marital status, education, gross yearly income, smoking, illicit substance and alcohol use in pregnancy); psychiatric correlates (i.e., severity of anxiety/depressive symptoms in as measured by SCL-5 at gestation week 17 and 30, self-reported lifetime history of major depression (LTH of MD), history of depression/anxiety before and/or in early pregnancy, co-medication in early pregnancy with antipsychotics, analgesic opioids, and anxiolytics/sedatives; paternal characteristics (e.g., age, paternal education, and anxiety/depressive symptoms by the SCL-8); infant factors (birthweight and prematurity, defined as gestational length <37 weeks). See Supplementary Material 1 for detail.
Data analysisTo estimate associations, we fitted crude and weighted generalized linear (GLM) and modified Poisson models with robust standard errors, for numeric and binary outcomes, respectively. To control for measured confounding, we used overlap weights based on the propensity score.27 These were estimated using logistic regression models for antidepressant initiation, overall and by timing, compared to unmedicated in the same time window postpartum, given the set of measured confounders. To minimize the risk of immortal time bias, mid-postpartum antidepressant initiators functioned as controls, i.e., were part of the unmedicated group, for the early postpartum initiation window. In the analysis of mid-postpartum antidepressant initiation, early antidepressant initiators were censored. Results are presented as crude and weighted mean differences, risk ratios (RR) and risk differences with corresponding 95% Confidence Intervals (CI). Moreover, we conducted stratified analyses in women with lowest (i.e., having greater polygenic score for positive antidepressant response) and highest (i.e., with lower polygenic score) tertial of the corresponding PGS. In addition, we adjusted the main weighted models for this PGS z-score.
Up to 5–20% of observations had missing values in at least one of the confounders or in one of the items composing the outcome measures. Under the assumption that data were missing at random, we conducted multiple imputation of missing values using chained imputation (20 replications).28 The multiple imputation model included the exposure variable, maternal and paternal covariates, and other auxiliary variables.
We conducted a series of sensitivity and sub-analyses. To examine associations by course and clinical type of psychiatric history, we restricted the population to women (i) with depression/anxiety only pre-pregnancy and (ii) with these disorders both pre- and during early pregnancy. We also limited the population to women with a history of (i) depression, (ii) anxiety, or (iii) any eating disorder. To understand the role of antidepressant treatment pre-pregnancy, we further adjusted and stratified our weighted models by this factor. In the subset with MoBa Genetics data, we adjusted weighted models for PGS for major depression and anxiety (z-scores), along with genotyping batch groups (n = 6) and the first five principal components to account for genetic substructure. Supplementary material 1 provides details on PGS calculation and sensitivity analyses. All analyses were performed in Stata/MP 18.
This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. MoBa was established with approval from the Norwegian Data Protection Agency and the Regional Committees for Medical and Health Research Ethics, in accordance with the Norwegian Health Registry Act.
ResultsWe included 1146 and 911 women at 6 and 18 months postpartum, respectively (Fig. 1), whereof 98.1% participated with one pregnancy/delivery. Of the 1146 women, 154 (13.4%) initiated antidepressants in the 0–6 months postpartum, and 992 (86.6%) had new mental illness episode within the same period but no antidepressant initiation. Of the 154 antidepressant initiators, 57 (37.0%) started in the first 0–3 months, while 97 (63.0%) did so in the 4–6 months postpartum. Most early initiators (48/57, 84.2%) continued treatment until 6 months postpartum. SSRIs were the antidepressant class most often initiated (131/154, 85.1%), followed by SNRIs (11/154, 7.1 %).
Abbreviations: MoBa=Norwegian Mother, Father and Child Cohort Study; MBRN=Medical Birth Registry of Norway; Q1=questionnaire 1 in MoBa; Q3=questionnaire 3 in MoBa; Q4= questionnaire 4 in MoBa; Q5= questionnaire 5 in MoBa.
Table 1 shows the baseline characteristics of the study population. Antidepressant initiators had more often pre-pregnancy psychiatric history than their peers, whereas the proportion with depression/anxiety in early pregnancy, and the severity of depressive and anxiety symptoms in early and late pregnancy, were comparable between the groups (Table 1). After overlap weighting, there was satisfactory balance (Standardized mean difference <0.1) in the distribution of confounders between the exposure groups (Figures S2-S3).
Baseline sociodemographic, lifestyle and health-related characteristics of the study population at six months postpartum by antidepressants initiation status. Data are presented as n (%) unless otherwise indicated.
| Antidepressant initiation, anytime postpartum (0–6 months) (n = 154) | Unmedicated mental illness postpartum (0–6 months) (n = 992) | |
|---|---|---|
| Pregnancy characteristics | ||
| BMI at conception; mean (SD) | 24.5 (5.1) | 24.1 (4.5) |
| Primiparous (yes) | 73 (47.4) | 530 (53.4) |
| Weight gain in pregnancy (kg); mean (SD) | ||
| Until GW 17 | 3.1 (3.5) | 2.9 (3.6) |
| From GW 17 to 30 | 5.9 (3.1) | 6.5 (3.3) |
| Maternal sociodemographic and lifestyle characteristics | ||
| Country of birth (Norway) | 142 (92.2) | 873 (88.0) |
| Missing | <5 | 14 (1.4) |
| Age groups (in years) | ||
| <24 | 24 (15.6) | 147 (14.8) |
| 25–35 | 104 (67.5) | 728 (73.4) |
| >35 | 26 (16.8) | 117 (11.7) |
| Married/Cohabitant (yes) | 141 (91.5) | 920 (92.7) |
| Educational level | ||
| University/College | 75 (48.7) | 530 (53.4) |
| Lower than University/College | 78 (50.6) | 455 (45.9) |
| Missing | <5 | 7 (0.7) |
| Grossly yearly incomea | ||
| Average | 95 (61.7) | 609 (61.4) |
| Low | 45 (29.2) | 285 (28.7) |
| High | 11 (7.1) | 67 (6.8) |
| Missing | <5 | 31 (3.1) |
| Alcohol use during pregnancy (yes) | 11 (7.1) | 121 (12.2) |
| Missing | <5 | 21 (2.1) |
| Illicit substances during pregnancy (yes) | <5 | 19 (1.9) |
| Smoking during pregnancy (yes) | ||
| Yes | 28 (18.1) | 171 (11.8) |
| Stopped during pregnancy | 15 (9.7) | 132 (13.3) |
| Missing | <5 | 12 (1.2) |
| Psychiatric and health-related characteristics | ||
| Antidepressant before pregnancy (yes) | 52 (33.8) | 103 (10.4) |
| Antidepressant in early pregnancy (yes) | 34 (22.1) | 67 (6.8) |
| Co-medication in early pregnancy (yes) | ||
| Antipsychotic use | <5 | 13 (1.3) |
| Anxiolytic use | <5 | 15 (1.5) |
| Opioid use | <5 | 28 (2.8) |
| SCL-5 at GW 17; mean (SD) | 1.7 (0.6) | 1.7 (0.6) |
| SCL-5 at GW 30; mean (SD) | 1.6 (0.5) | 1.7 (0.6) |
| LTH of MD (yes) | 56 (36.7) | 281 (28.3) |
| Missing | 5 (3.3) | 24 (2.4) |
| Depression before pregnancy (yes) | 99 (64.3) | 571 (57.6) |
| Depression in early pregnancy (yes) | 27 (17.5) | 188 (18.9) |
| Anxiety before pregnancy (yes) | 74 (48.1) | 336 (33.9) |
| Anxiety in early pregnancy (yes) | 20 (12.9) | 129 (13.0) |
| Eating disorder prior to pregnancy | ||
| Anorexia nervosa | 0 | 8 (0.8) |
| Bulimia nervosa | 10 (6.5) | 57 (5.7) |
| Binge eating disorder | 10 (6.5) | 165 (16.6) |
| Missing | 39 (25.3) | 170 (17.1) |
| Eating disorder in early pregnancy | ||
| Bulimia nervosa | 0 | 23 (2.3) |
| Binge eating disorder | 17 (11.0) | 191 (19.2) |
| Missing | 32 (20.8) | 152 (15.3) |
| Paternal factors | ||
| Age (in years) | ||
| <25 | 11 (7.1) | 76 (7.7) |
| 25–39 | 124 (80.5) | 795 (80.1) |
| 40–49 | 18 (11.7) | 109 (11.0) |
| >49 | 0 | <5 |
| Educational level | ||
| University/College | 42 (27.3) | 354 (35.7) |
| Lower than University/College | 68 (44.2 | 419 (42.2 |
| Missing | 44 (28.6) | 219 (22.1) |
| SCL-8 at GW 17; mean (SD) | 1.2 (0.3) | 1.2 (0.4) |
| Infant characteristics | ||
| Breastfeeding (months) | ||
| Up to 3 months postpartum | 61 (39.6) | 150 (15.1) |
| From 4 to 7 months postpartum | 75 (48.7) | 763 (76.9) |
| Sex (Male) | 82 (53.2) | 495 (50.0) |
| Birth weight (in grams); mean (SD) | 3582 (596) | 3546 (541) |
| Prematurity (yes) | 6 (3.9) | 58 (5.8) |
Average indicates income approximately between 17,501–46,800 USD; Low indicates income ≤ 17,500 USD; High indicates income ≥ 46,801 USD. bEating disorder not other specified and eating disorder not other specified, purging type, are not shown due to very low counts across groups. Abbreviation: SD=standard deviation; BMI=Body mass index; SCL=Hopkins symptom checklist; GW=Gestational week; LTH of MD= Lifetime History of Major Depression.
Antidepressant initiation in the 0–3 months postpartum was associated with lower severity of depressive/anxiety symptoms at 6 months post-delivery (weighted b= −0.23, 95% CI = −0.39, −0.06) but the association did not persist at 18 months (weighted b= −0.07, 95% CI = −0.28, 0.14) (Fig. 2, Supplementary Table S2). The association at 6 months was driven by a reduction of the depressive symptoms specifically (weighted b= −0.28, 95% CI = −0.48, −0.09). At 18 months, depressive symptoms were slightly lower with antidepressant initiation in mid-postpartum, albeit the 95% CI included the null value. There was no or minimal reduction of anxiety symptoms based on the EPDS and the SCL subscales with antidepressant initiation, and likewise no association with self-esteem and satisfaction with life at both time poin
Association between antidepressant initiation postpartum, overall and by timing, and severity of mental health outcomes at 6 and 18 months postpartum, in the weighted models.
The “SCL total” measures combined symptom severity of depression/anxiety; the two subscales of the SCL measure specifically anxiety and depressive symptom severity. Any time postpartum indicates antidepressant initiation within 0–6 months postpartum; mid-postpartum indicates initiation in months 4–6; early postpartum indicates initiation in months 0–3.
The proportion of women experiencing active, clinically relevant symptoms of depressive/anxiety, using a strict definition as SCL-8 score ≥2.0, was elevated across exposure groups at 6 months postdelivery (22.8%−43.9%) (Figure S4). Antidepressant initiators in the 0–3 months postpartum had a 44% reduced risk of having strictly defined, active depressive/anxiety symptoms, relative to unmedicated subjects, but only at 6 months post-delivery (weighted RR= 0.56, 95% CI = 0.35, 0.91; risk difference= −0.57) (Fig. 3, Supplementary Table S3). The magnitude of this association decreased when using the broader outcome definition. There was no strong evidence for such an association with antidepressant initiation in months 4–6 at both 6 and 18 months since childbirth. Antidepressant initiation postpartum, regardless of timing of treatment start, was not associated with low self-esteem or active anxiety symptoms based on the EPDS subscale (Fig. 3, Supplementary Table S3).
Association between antidepressant initiation postpartum, overall and by timing, and active, clinically relevant mental health outcomes at 6 and 18 months postpartum, in the weighted models.
Any time postpartum indicates antidepressant initiation within 0–6 months postpartum; mid-postpartum indicates initiation in months 4–6; early postpartum indicates initiation in months 0–3.
The association measures between depressive/anxiety symptom outcomes and true antidepressant initiation within 0–6 months postpartum were of slightly greater effect size (about 10%) than the main results; for early initiation, the change in effect size was about 40–45% (Supplementary Table S4-S5).
Upon stratification by residual depressive/anxiety symptom at gestational week 30, associations with antidepressant initiation in the first 0–3 postpartum months were evident only among women without such symptoms (Supplementary Table S6). Associations stratified by clinical type or illness course (Supplementary Figures S5–S8) did not materially change our main results, except for women with a history of eating disorders showed no clinical benefit from early initiation. Additional adjustment for antidepressant treatment in the 6 months pre-pregnancy did not alter results (Supplementary Figure S9). Upon stratification at gestation week 30, reduced risk of active depressive/anxiety symptoms appeared only among women treated pre-pregnancy (Supplementary Figure S10).
In the sub-population with genetic data (n = 462), adjusting for the PGS z-score for antidepressant response, anxiety, and depression did not materially change the 6-month results (Fig. 4, Supplementary Tables S7–S8). Among women with higher polygenic scores for positive antidepressant response (lowest tertile), antidepressant initiation in the first 0–3 postpartum months was more strongly associated with lower depressive/anxiety symptoms at 6 months (weighted b = –0.50, 95% CI: –0.84, –0.16), particularly for depressive symptoms (b = –0.69, 95% CI: –1.03, –0.34), with effects persisting at 18 months (b = –0.77, 95% CI: –1.24, –0.29). No such associations appeared among women with lower polygenic scores (Fig. 4, Supplementary Tables S7–S8). See Supplementary Material 1 for additional sensitivity analyses.
Weighted associations between antidepressant initiation postpartum and severity of anxiety and depression at 6 (Panel A) and 18 (Panel B) months postpartum, adjusted and stratified by polygenic score for antidepressant response.
This study reports the clinical benefit of antidepressants initiation at different timings during the first half of the postpartum year, on short- and long-term mental health outcomes in women with psychiatric history before and/or during early pregnancy. Antidepressant initiation within the first three months postpartum was associated with lower depression/anxiety symptom severity at 6 months, but not at 18 months, compared with unmedicated women experiencing a new episode. The short-term improvement was mainly within the depressive clinical subtype, with no or minimal associations for anxiety, self-esteem, or life satisfaction. Upon dichotomization of the outcome measure, our weighted risk difference of having active, clinically relevant depression/anxiety symptoms at 6 months post-delivery with antidepressant initiation within 0–3 months postpartum was as high as 57 per 100. Exposure misclassification could have underestimated these estimates by 10–40%, leading to an unaltered inference. These findings, including the magnitude of the association measures, are consistent with previous clinical trials showing depression remission and lower depression severity with antidepressant intervention in the first months postdelivery.9,29 Because the first months postpartum are a period of increased risk for both new and recurrent episodes of psychiatric hospital admission,6 timely treatment with antidepressant is critical for safeguarding maternal mental health and mother-child well-being, especially since these medications are compatible with breastfeeding.30
Our risk reduction for active, clinically relevant depression/anxiety symptoms at 6 months with early antidepressant initiation was notable among women treated only in the 6 months pre-pregnancy and without residual symptoms at the end of gestation. Although we could not determine whether postpartum reinitiation reflected prophylaxis or emerging symptoms, our findings highlight the need for larger studies evaluating the effectiveness of prophylactic postpartum antidepressant use in high-risk women.
We found no substantial association between antidepressant initiation at 4–6 months postpartum and short- or long-term outcomes. Some factors may explain these results. First, our mid-postpartum initiation window was broad, and many women may have started treatment close to 6 months, giving insufficient time for antidepressants to exert effects. Second, we also could not determine whether treatment was delayed due to breastfeeding or how such delays might have influenced antidepressant response. GWAS and epidemiological data have documented that depression with onset in the first 2 versus 3–12 months postpartum constitute two distinct phenotypes, with variation in severity and etiological factors.31 Yet, whether antidepressants have differential effectiveness for treatment of these distinct phenotypes remains untestable in the current study.
There is a continued imperative to better understand whether, and for whom, antidepressant treatment is effective for postnatal depression and other mental illnesses.9 Although our polygenic score (PGS) analyses were exploratory and based on a relatively small sample, reduction in depressive and anxiety symptom severity following antidepressant initiation within the first three months postpartum was observed only among women with higher genetic liability toward positive antidepressant response. The observed effect size was stronger than in the main analysis, with clinical benefits persisting up to 18 months postdelivery. Associations were consistent when using PGS for maternal depression and anxiety. However, these results should be interpreted cautiously, as our PGS for antidepressant response included only 27 SNPs and captured limited inheritance. Still, the findings support the need for larger studies on personalized antidepressant treatment in perinatal psychiatry. This study has methodological limitations that need to be considered when interpreting the findings. Furthermore, this study was based on data from 1999 to 2008. Since that time, screening, recognition, and management of depression during pregnancy have improved substantially, with expanded treatment options and updated clinical guidance. Therefore, the findings observed in our study may not fully reflect current practice. The population was small in specific sensitivity analyses, and genetic data were available only for a subsample, though this remains a notable strength. Women reported on their antidepressant use and mental health outcomes in the same questionnaire at 6 months since childbirth, which can introduce social desirability and response biases. However, we quantified the impact of exposure misclassification on our associations, and these moved further away from the null, leading to an unaltered inference. The study has no information on antidepressant dose, use at 18 months postpartum, granular duration of treatment, and specific timing of onset of the mental illness within the 0–6 months postpartum. Lower impact at 18 months compared to 6 months, however, could be attributed to potential discontinuation or treatment dropouts after few months of treatment which is relatively common among antidepressant initiators. Our outcomes were self-reported by women and not based on clinical interview; however, the psychometric instruments used had satisfactory reliability in this study, and have been previously validated.22–24 By using this outcome set, we could examine the effectiveness of antidepressants on the symptom severity spectrum, by sub-domain level, and on clinically relevant binary outcomes. Although baseline symptom severity during pregnancy was included in the weighting model and balanced after adjustment, severity at the time of treatment initiation was not directly measured. Therefore, residual confounding by severity cannot be fully excluded. The MoBa study has a low response rate (41%), with a possible self-selection of the healthiest women into the cohort.14 Although association measures have been shown to be valid in MoBa in relation to immediate birth outcomes,32 the impact of selection bias on the postpartum maternal outcomes cannot be excluded.
In women with psychiatric history before and/or during early pregnancy, initiating antidepressants in the first 3 months postpartum was linked to lower depressive/anxiety symptoms at 6 months. The lower risk was more evident and long-lasting in women with high polygenic score towards antidepressant response. Further research is needed to clarify differences in effectiveness for early- and late-onset postpartum mental illness.
ContributorsAL led this study and contributed to data acquisition; AL and ZFC conducted the statistical analyses and drafted the manuscript. SR conducted the analyses of genetic data. NTHT verified analytical scripts. All authors contributed to clinical interpretation of the study results, revised the manuscript draft critically for important intellectual content, and approved the final version of the manuscript.
Data availabilityData from MoBa and the Medical Birth Registry of Norway are protected under Norwegian law and cannot be publicly shared. Access requires approval from the Norwegian Institute of Public Health, the Regional Committees for Medical and Health Research Ethics (REC), and compliance with the EU General Data Protection Regulation (GDPR). Data access requests can be submitted through helsedata.no.
Ethical considerationsThe present study was approved by the Regional Committees for Medical and Health Research Ethics (Ref: 63,566/REK Sør-Øst) and the Data Protection Services (DPIA: 672,954).
FundingNhung Trinh and Angela Lupattelli were supported by the Norwegian Research Council (grant 288,696). The Norwegian Mother, Father and Child Cohort Study is supported by the Norwegian Ministry of Health and Care Services and the Ministry of Education and Research. The funder had no role in the study design; in the collection, analysis, or interpretation of data; in the writing of the report; or in the decision to submit the article for publication.
The authors declare no conflict of interest.
We thank Prof. Eivind Ystrøm and Dr. Rosa Catherine Gillespie Cheesman for support with the MoBa Genetics quality-control pipeline, and we are grateful to all families participating in the Norwegian Mother, Father and Child Cohort Study (MoBa). We also acknowledge the Norwegian Institute of Public Health for generating genomic data. Veerle Bergink was funded by the U.S. National Institute of Mental Health (R01MH122869). This study forms part of the HARVEST collaboration (Research Council of Norway, (229624). Genotype data from the NORMENT Centre were supported by the Research Council of Norway (223273), Southeast Norway Health Authorities, and Stiftelsen Kristian Gerhard Jebsen. Additional genotype data and quality control were provided by the Center for Diabetes Research, University of Bergen, funded by the ERC Advanced Grant SELECTionPREDISPOSED, the Trond Mohn Foundation, the Research Council of Norway, the Novo Nordisk Foundation, the University of Bergen, and Western Norway Health Authorities.
