Consequences of Maternal Antidepressant and Anticonvulsant Use During Pregnancy: Is Postnatal Inpatient Monitoring Required? – A Retrospective Cohort Study

Abstract

Introduction

Prenatal exposure to antidepressants and anticonvulsants is steadily increasing in parallel with rising prescription rates. Our study focused on potential postnatal adjustment disorders in newborns and their appropriate monitoring after birth.

Objective

To descriptively assess clinically relevant complications in neonates after intrauterine exposure to antidepressants and/or anticonvulsants and to evaluate the need for routine postnatal monitoring.

Methods

We conducted a retrospective single-center cohort study at the University Hospital Erlangen (01/2021–12/2023). The study included 120 neonates whose mothers received antidepressants, anticonvulsants, or both during pregnancy. The primary endpoint was a clinically relevant respiratory event (apnea, dyspnea/desaturation, or need for ventilatory support). An exploratory multivariable logistic regression with the primary endpoint as the outcome was performed (main analysis: gestational age, maternal age, cesarean delivery; sensitivity analysis additionally including exposure group).

Results

The primary endpoint occurred in 31/120 neonates (25.8%). In the main multivariable analysis, gestational age, maternal age, and cesarean delivery showed no statistically significant association with the primary endpoint; in a sensitivity analysis additionally including exposure group, no significant between-group differences were observed. Numeric time-to-event data were available for 22 respiratory events (median 32.5 h; mean 42.0 h); 16/22 (72.7%) occurred within 48 h and 6/22 (27.3%) after 48 h. Hypoglycemia was documented in 15/120 neonates (12.5%) without significant differences between exposure groups.

Conclusion

Clinically relevant respiratory events were observed in this monitored cohort of exposed neonates. An initial monitoring period of at least 48 h appears to represent a pragmatic minimum standard. In the presence of clinical abnormalities, risk-adapted prolonged monitoring may be appropriate.

Introduction

In recent years, antidepressant prescription rates have gone up, with women of childbearing age being particularly affected 1 2 3 . In 2025, a five-fold increase was shown by the data from Denmark, with 3.9% of pregnant women there taking antidepressants in 2023 4 . Pregnancy and the perinatal period are considered to be the periods associated with an increased risk of depression, bipolar disorder and anxiety disorders, with reported prevalence rates ranging from 6.5% to 12.9%, depending on the stage of pregnancy and the first postpartum year 5 . An international meta-analysis involving 14072251 pregnancies identified selective serotonin reuptake inhibitors (SSRIs) as the most frequently used antidepressants during pregnancy. The prevalence worldwide was 3.0%, although there were significant regional differences: 5.5% in North America, 1.6% in Europe and 1.3% in Australasia 3 .

The fact that it is often necessary to continue antidepressant treatment during pregnancy and beyond creates an area of conflict: on the one hand, the treatment of maternal psychiatric disorders is clinically relevant, as untreated disorders are themselves associated with maternal and perinatal risks, such as preterm birth 6 , as well as maternal mortality 7 . On the other hand, there is consistent evidence of an association between intrauterine exposure and poor neonatal adaptation syndrome (PNA), including increased risks of respiratory failure, hypoglycaemia and seizures 8 9 10 11 .

With regard to potential long-term neurocognitive effects, the current evidence shows no indication of an increased risk: in a cohort of more than three million pregnancies, Suarez et al. (2022) found no association with autism spectrum disorders, ADHD or intellectual disability 12 .

Compared with international guidelines, the current version (2022) of the National Clinical Practice Guideline on Unipolar Depression contains only limited specific recommendations on the treatment of depressive disorders during pregnancy 13 . Recommendations on non-pharmacological treatment options can be derived from the previous 2017 guideline: Pregnant women with depression should be offered psychotherapy; prophylactic psychotherapeutic intervention may also be considered for women at increased risk of depression. Physical exercise may be recommended to alleviate symptoms; in cases of severe depression, electroconvulsive therapy may be considered as a complementary treatment 14 .

An integrative approach that takes individualised account of psychotherapeutic and pharmacological options has the potential to reduce neonatal medication exposure.

Alongside antidepressants, anticonvulsants are also frequently prescribed and are often essential medications for expectant mothers, raising similar questions regarding neonatal safety 15 .

To date, there are no further guidelines available in Germany on caring for these newborns; the 2017 guideline on unipolar depression merely recommends that, in cases of intrauterine antidepressant exposure, the mother should give birth at a clinic with an attached neonatal unit 14 . The current 2022 edition does not contain any recommendations on postnatal management 13 .

Only international treatment guidelines provide more specific recommendations for action, according to which newborns should be monitored as inpatients. One example is the Clinical Excellence Commission of New South Wales, Australia, which recommends 24-hour monitoring of antidepressant-exposed newborns 16 . A guideline from the UK recommends monitoring for 48 hours 17 . The NWC Perinatal Mental Health Network also recommends monitoring newborns exposed to both antidepressants and anticonvulsants 18 . The documents mentioned were published primarily with a focus on consensus and patient care, rather than as evidence-graded guidelines. Another guideline from the UK recommends that newborns exposed to psychoactive medications in utero, including antidepressants and anticonvulsants, should be monitored for several weeks for toxic effects of the medications and withdrawal symptoms 19 . The latter document is a national evidence-based guideline with formal methodology. In view of the nature of the evidence cited, the recommendations are discussed as clinically plausible courses of action, with causal basis not yet conclusively established.

At Erlangen University Hospital, affected newborns are transferred to the neonatal unit for postnatal monitoring in accordance with internal standards. They are generally discharged after the second newborn examination (U2) or following a 48-hour observation period without any complications. This procedure reliably identifies any respiratory and metabolic problems that the babies may suffer from and allows for immediate supportive intervention; however, it also requires that the mother and baby are separated postnatally.

Against this background, the aim of this retrospective, single-centre cohort study was to descriptively assess morbidity and intervention rates among exposed newborns and to describe the temporal patterns of clinical events.

Methods

Study design and population

A retrospective, pseudonymised analysis was carried out at Erlangen University Hospital for the period of 1 January 2021 to 31 December 2023. The aim was to identify neonatal patients (m/f/nb) at Erlangen University Hospital [UKER] who had been monitored as inpatients due to intrauterine exposure to antidepressants or anticonvulsants. They make up our study population of 120 children.

To provide contextual background, the nationwide cohort of the Institute for Quality Assurance and Transparency in Health Care (IQTIG) from the years 2016 to 2019 was used. The IQTIG analyses only cover the allocation rates of diagnoses relating to child morbidity (ICD-10), including P22 and P28 diagnoses (P22.0, P22.1, P22.8, P22.9, P28.-), up to the year 2019, so we selected the data for the years 2016–2019.

Inclusion and exclusion criteria

Our study population included all newborns from the years specified, provided that intrauterine exposure to antidepressants, anticonvulsants or a combination of both substance classes had been documented.

Exclusion criteria:

• Mode of delivery: operative delivery under general anaesthesia with intubation

• Discontinuation of maternal medication prior to birth

• Additional maternal intake of other substances that could potentially affect postnatal adaptation

Data collection

Data was collected through a systematic review of electronic patient records using the Soarian Clinicals (Cerner Health Services) and Integrated Care Manager (ICM by Dräger Medical GmbH, version 14.11) systems. The following variables were recorded:

• Gestational age

• Mode of delivery

• Sex

• Maternal age

• Medication use

• APGAR scores (1-, 5- and 10-minute scores)

• pH of umbilical cord blood

• Presence of malformations

• Respiratory support (CPAP or invasive ventilation) and its duration

• Clinical abnormalities (e.g. apnoea, drops in oxygen saturation, hypoglycaemia)

• ICD-10 diagnoses assigned

Data was collected using standardised documentation forms in a tabular format.

Cohort characterisation

To characterise the study population, the data were compared with the IQTIG results for the years 2016–2019. They were stratified in accordance with:

• Gestational age: 28–31, 32–36, 37–41, > 41 weeks of gestation

• Birth weight: < 500 g, 500–759 g, 750–999 g, 1000–1499 g, 1500–1999 g, 2000–2499 g, 2500–2999 g, 3000–3999 g, 4000–4499 g, ≥ 4500 g

• Mode of delivery: Spontaneous vaginal delivery, planned or emergency caesarean section, operative delivery (vacuum extraction or forceps)

• Maternal age: < 18, 18–29, 30–34, 35–39, ≥ 40 years

Documented medication intake was grouped as follows:

• Antidepressants: SSRIs, SNRIs, tricyclics, tetracyclics

• Anticonvulsants: 1st generation (carbamazepine, phenytoin, valproate), lamotrigine/levetiracetam, 2nd generation excluding lamotrigine/levetiracetam

• Psychotropic drugs: quetiapine, lithium, clozapine

Data was also grouped with regard to monotherapy and combination therapies.

To assess postnatal adaptation, the APGAR scores at 5 and 10 minutes were compared between the two groups, with an APGAR score of < 7 being defined as pathological.

The diagnoses were recorded in accordance with the ICD-10 codes.

Statistical analyses

Apnoea, dyspnoea, drops in oxygen saturation or respiratory episodes requiring ventilation beyond the stay in the delivery room were classified as clinically relevant respiratory events. Short-term drops in saturation were only considered as events if they necessitated a clinical response or intervention by medical staff, or if they were documented as clinically relevant during the clinical course. Isolated, spontaneously reversible pulse oximetry events with no documented clinical consequences were not counted as a primary end point.

Respiratory end points are consistently referred to in the text as ‘respiratory events’; where diagnostic data is concerned, this is explicitly stated as ‘ICD-coded respiratory diagnoses (P22/P28)’.

Exposure was classified into four groups:

• Antidepressant monotherapy

• Anticonvulsant monotherapy

• Other monotherapy (lithium, quetiapine or clozapine)

• Combination medication

Descriptive variables were given as absolute and relative frequencies. To assess the association between clinical risk factors and the primary end point, a multivariable logistic regression analysis was performed, with fulfilment of the primary end point (apnoea/dyspnoea/drop in oxygen saturation/ventilation) as the dependent variable. The following covariates were taken into account: gestational age (continuous, in weeks), maternal age (continuous, in years) and mode of delivery (caesarean section yes/no, excluding procedures under general anaesthesia). As a sensitivity analysis, a multivariable logistic regression was also performed, in which the exposure group was also taken into account. The subgroup receiving antidepressant monotherapy was used as the reference category.

All documented time values were included in the analysis of the time to the last documented respiratory event. The data were also categorised into predefined time intervals (0–6 h, 6–24 h, 24–48 h, > 48 h). Due to small sample sizes and non-normally distributed data, a Kruskal-Wallis test was primarily used to compare the time distributions between exposure groups. As part of an exploratory analysis, a pairwise Mann-Whitney U test was also reported (antidepressant monotherapy vs. anticonvulsant monotherapy) as a sensitivity analysis. In addition, the influence of extreme individual values was examined for sensitivity analyses.

The results are presented as odds ratios (OR) with 95% confidence intervals and p values. A complete-case analysis was conducted. Due to small sample sizes in the individual subgroups, the analysis should be interpreted as exploratory and hypothesis-generating.

Ethics

This study was approved by the Ethics Committee of the Institutional Review Board of the Friedrich-Alexander University of Erlangen–Nuremberg (No. 24–229-Br, date: 02 July 2024)

Results

In-utero medication exposure

In our study population (n = 120), 101 newborns were assigned to a monotherapeutic exposure group. Among these, the most common monotherapy was with SSRIs (n = 60), followed by lamotrigine/levetiracetam (n = 19). Other medications included SSNRIs, tricyclic antidepressants and selected psychotropic drugs such as quetiapine, lithium and clozapine (see Table 1 ).

Table 1 Distribution of monotherapies used by drug class.

Total monotherapy	Number (n)	Percentage (%)
Total	101	100
Of which
SSRI	60	59.41
SSNRI	11	10.89
Tricyclics	4	3.96
Lamotrigine/levetiracetam	19	18.81
Quetiapine	4	3.96
Lithium	2	1.98
Clozapine	1	0.99

A total of 19 babies were subject to combined in-utero exposure. The most frequent combinations were SSRI + anticonvulsant and SSRI + antipsychotic. Individual additional combinations are listed in Table 2 .

Table 2 Overview of combination medications in the study population.

Combination	Drug class(es)	Cases (n)
Lamotrigine/topiramate	Anticonvulsant (1st + 2nd generation)	2
Lamotrigine/levetiracetam	Anticonvulsant (2nd generation)	6
Levetiracetam/oxcarbazepine	Anticonvulsant (2nd generation)	1
Levetiracetam/sertraline	Anticonvulsant (2nd generation) + SSRI	1
Amitriptyline/sertraline	Tricyclic antidepressant + SSRI	1
Amitriptyline/venlafaxine	Tricyclic antidepressant + SSNRI	1
Gabapentin/escitalopram	Anticonvulsant + SSRI	1
Quetiapine/escitalopram	Psychotropic medication + SSRI	1
Mirtazapine/sertraline	Psychotropic medication + SSRI	1
Quetiapine/sertraline	Psychotropic medication + SSRI	1
Mirtazapine/sertraline	Tetracyclic antidepressant + SSRI	1
Sertraline/quetiapine/lamotrigine	SSRI + psychotropic drug + anticonvulsant	1
Quetiapine/amitriptyline/duloxetine	Psychotropic drugs + tricyclic antidepressant + SSNRI	1

Respiratory outcomes

The primary end point was defined as an episode of apnoea, dyspnoea/desaturation or a respiratory episode requiring ventilation beyond the time in the delivery room. In the overall cohort, the primary end point occurred in 31 out of 120 newborns (25.8%).

Descriptively, the following rates of the primary end point were observed in the exposure groups: antidepressant monotherapy 17/75 (22.7%), anticonvulsant monotherapy 7/19 (36.8%), combination exposure 4/19 (21.1%) and other monotherapy 3/7 (42.9%).

An exploratory multivariable logistic regression was carried out to assess the association between clinical risk factors and the primary end point (n = 116 complete data sets). The primary end point served as the dependent variable; gestational age (in weeks), maternal age (in years) and caesarean section (yes/no) were included as covariates. No statistically significant association was found:

• Gestational age: OR 0.93 (95% CI 0.74–1.16; p = 0.515),

• maternal age: OR 1.02 (95% CI 0.94–1.10; p = 0.639),

• caesarean section: OR 1.22 (95% CI 0.52–2.87; p = 0.652).

As a sensitivity analysis, a multivariable logistic regression was also carried out, with the defined primary end point as the dependent variable and additional consideration of the exposure group (reference group: antidepressant monotherapy; n = 116). No statistically significant correlation was found:

• Gestational age OR 0.91 (95% CI 0.72–1.14; p = 0.409),

• maternal age OR 1.02 (95% CI 0.95–1.11; p = 0.549),

• caesarean section OR 1.26 (95% CI 0.52–3.00; p = 0.610),

• anticonvulsant monotherapy vs. antidepressant monotherapy OR 2.09 (95% CI 0.70–6.26; p = 0.188),

• Combination vs. antidepressant monotherapy OR 0.94 (95% CI 0.27–3.26; p = 0.918),

• other monotherapy vs. antidepressant monotherapy OR 3.00 (95% CI 0.59–15.30; p = 0.185).

Given the limited number of cases and the wide confidence intervals, both analyses should be interpreted as exploratory and hypothesis-generating. The comparison of drug groups in the sensitivity analysis does not allow for a robust differentiation between exposure groups.

All numerically documented time values were taken into account for the time analysis of respiratory events. Of the 31 cases with the above-mentioned primary end point, numerical time data on the most recently documented occurrence were available in 22 cases. The median was 32.50 h and the mean was 41.99 h after birth (Q1: 12.48 h; Q3: 50.87 h; range: 0.30–197.0 h). The time window distribution is as follows: 0–6 h: 4/22 (18.2%), 6–24 h: 6/22 (27.3%), 24–48 h: 6/22 (27.3%), > 48 h: 6/22 (27.3%).

The subgroup analysis of the time to the last documented respiratory event revealed the following median values: antidepressant monotherapy 13.0 h (n = 13), anticonvulsant monotherapy 51.83 h (n = 5), combination therapy 24.16 h (n = 2), other monotherapy 65.53 h (n = 2). The global comparison between the groups was not statistically significant in the Kruskal-Wallis test (p = 0.097). After excluding the outlier of 197 h, a sensitivity analysis revealed a numerical shift (Kruskal-Wallis p = 0.0356; exploratory analysis: antidepressant monotherapy vs. anticonvulsant monotherapy in the Mann-Whitney U test p = 0.0176), which, given the small subgroup sizes, should be interpreted as hypothesis-generating.

Hypoglycaemia

Of a total of 120 newborns examined, hypoglycaemia occurred in 15 children (12.5%). The temporal analysis revealed a median of 22.0 h and a mean of 25.61 h after birth. The time-window distribution of these 15 numerical values was as follows: 0–6 h: 0/15 (0.0%), 6–24 h: 8/15 (53.3%), 24–48 h: 7/15 (46.7%), > 48 h: 0/15 (0.0%).

The subgroup analysis revealed no statistically significant differences between the exposure groups (antidepressant monotherapy: 11/75 [14.7%], anticonvulsant monotherapy: 1/19 [5.3%], combination therapy: 2/19 [10.5%], other monotherapy: 1/7 [14.3%]; overall comparison p = 0.724). No significant difference was observed in the pairwise comparison of antidepressant monotherapy vs. anticonvulsant monotherapy (Fisher’s test, p = 0.448).

Due to the small number of cases in the subgroups, the statistical power of these comparisons is limited; the results should therefore be interpreted as exploratory and hypothesis-generating.

Contextual classification using the IQTIG cohort

Table 3 provides an overview of the differences in maternal age distribution, newborn sex ratios, modes of delivery and birth weight in the study population and the IQTIG cohort.

Table 3 Distribution of characteristics across the various groups.

	UKER [University of Erlangen–Nuremberg]		IQTIG
	Number (n)	Percentage (%)	Number (n)	Percentage (%)
Maternal age
< 18	0	0.00	15195	0.50
18–29	30	25.00	1180439	39.03
30–34	45	37.50	1086021	35.90
35–39	31	25.83	613365	20.28
≥ 40	12	10.00	129833	4.29
n.a.	2	1.67	0	0.00
Sex
M	57	47.50	1580282	51.25
F	63	52.50	1502638	48.73
n.a.	0	0.00	700	0.02
Type of delivery
Spontaneous	73	60.83	1886117	61.09
Caesarean section	43	35.83	980115	31.54
of which planned	22	18.33	212467	13.71
of which emergency	21	17.50	246558	15.91
of which unspecified	0	0.00	36948	2.38
Operative vaginal	4	3.33	212234	6.72
Other	0	0.00	5244	0.17
Birth weight
< 500 g	0	0.00	3865	0.13
500–759 g	0	0.00	8188	0.27
750–999 g	0	0.00	8618	0.28
1000–1499 g	0	0.00	22673	0.74
1500–1999 g	0	0.00	43118	1.30
2000–2499 g	8	6.67	127424	4.13
2500–2999 g	23	19.17	478242	15.51
3000–3999 g	79	65.83	2075915	67.32
4000–4499 g	9	7.50	277679	9.01
≥ 4500 g	1	0.83	37898	1.23
pH value
< 7.00	1	0.83	7765	0.25
7.00–7.09	4	3.33	60611	1.98
7.10–7.19	21	17.5	425230	13.91
≥ 7.20	84	70.0	2550610	83.41
n.a.	10	8.33	0	0.0

The maternal age distribution differed between the groups (chi-square test, p < 0.001). Whilst 25.0% of mothers in the UKER cohort were younger than 30, the corresponding proportion in the IQTIG comparison group was 39.0%. At 10.0%, the proportion of mothers aged ≥ 40 in the UKER cohort was more than twice as high as in the comparison group (4.3%).

With regard to newborn sex, the distribution was very similar: male newborns at the UKER: 47.5%, in the IQTIG group: 51.3%.

Likewise, the distribution of modes of delivery between the two cohorts is very similar: in the Erlangen cohort, we recorded 60.8% spontaneous vaginal deliveries and 35.8% caesarean sections; in the IQTIG group, the figures were 61.1% for spontaneous vaginal deliveries and 31.5% for caesarean sections.

The distribution of birth weights showed only minor differences between UKER and IQTIG. The majority of newborns weighed between 3000–3999 g (UKER: 65.8%; IQTIG: 67.3%).

The distribution of gestational age differed significantly (p < 0.001). In the UKER cohort, 84.2% of newborns were born between 37 and 41 weeks’ gestation (IQTIG: 91.0%), whilst a higher proportion were recorded as post-term (> 41 weeks’ gestation) (UKER: 7.5%; IQTIG: 0.5%). There were no preterm births at < 28 weeks’ gestation in the UKER study cohort (IQTIG: 0.6%). One case was recorded in the 29th week of gestation, eight cases in the 36th week.

Comparing the assigned diagnoses P22 (‘respiratory distress syndrome of the newborn’) and P28 (‘other respiratory disorders’) between the Erlangen cohort and the IQTIG cohort shows that in the nationwide group, 1.98% (n = 60998) received a diagnosis of P22 or P28, whilst in the UKER study cohort the prevalence was 34.17% (n = 41).

Table 4 shows the distribution of diagnoses by drug group within the overall study population.

Table 4 Distributions of P28 and P22 diagnoses in the study population.

Assigned diagnoses	P28		P22
	Number (n)	Percentage (%)	Number (n)	Percentage (%)
Total study population (n = 120)
Total	30	25.0	11	9.17
Antidepressant monotherapy	15	12.50	9	7.5
Anticonvulsant monotherapy	7	5.83	1	0.83
Full-term infants (n = 110)
Total	27	24.55	10	9.09
Antidepressant monotherapy	13	11.82	8	7.27
Anticonvulsant monotherapy	6	5.45	1	0.91

Analysis of the APGAR scores for the study population at University Hospital Erlangen showed a median 5-minute APGAR score of 10 with an interquartile range of 1; the 10-minute APGAR score also had a median of 10 with an interquartile range of 0. Compared with the distributions of 5-minute APGAR scores in the IQTIG cohort, 54.2% (IQTIG: 78.2%) of the study population had an APGAR score of 10, whilst 35.8% (IQTIG: 20.2%) had an APGAR score of 7, 8 or 9.

Analysis of the pH values in umbilical cord blood revealed a discrepancy between the UKER cohort and the IQTIG database, as shown in Table 3 : there was a higher proportion of perinatal acidosis in the UKER study group.

Discussion

The present results show that relevant respiratory events occurred in 25.8% of a monitored UKER cohort with intrauterine exposure.

The multivariable regression analysis to classify clinical risk factors (gestational age, maternal age, caesarean section) in relation to the primary end point revealed no statistically significant associations. As such, the respiratory events observed in our cohort could not be explained solely by the perinatal risk factors mentioned. Even in the sensitivity analysis, which also took the exposure groups into account, no significant differences were found between the drug groups after adjustment. This argues against robust group-specific effects in the present sample; however, given the limited number of cases and the study design, such effects cannot be ruled out.

With regard to the clinical evaluation of the current procedure at University Hospital Erlangen, which involves monitoring up to U2 (the second newborn screening) – and therefore for at least 48 hours – the recorded time data revealed the following: 72.7% of respiratory events occurred within 48 hours; 27.3% occurred after 48 hours. Whilst this supports 48-hour monitoring as an advisable minimum standard, it also demands individual risk adaptation. An approach involving extended monitoring in the event of clinical abnormalities or risk constellations therefore seems advisable.

The time data relating to respiratory events descriptively suggest group-specific patterns; however, due to the small subgroups, these are of limited reliability in inferential statistical terms.

When viewed in the context of the IQTIG reference, the UKER cohort shows higher rates of postnatal complications, particularly respiratory adaptation disorders. 25% received the diagnosis P28 (other respiratory disorders originating in the perinatal period) (IQTIG: 0.88%) and 9.17% P22 (other respiratory distress of newborn) (IQTIG: 1.10%).

The 5- and 10-minute APGAR scores were also lower in our cohort than in the nationwide comparison group (5 minutes: 5.00% vs. 1.26%; 10 minutes: 1.67% vs. 0.42%).

In addition to drug exposure, mode of delivery and gestational age are independent risk factors for neonatal respiratory adaptation disorders. Transient tachypnoea of the newborn (TTN) occurs with an incidence of up to 30% following elective caesarean section, due to delayed resorption of foetal lung fluid in the absence of labour-related stress 20 .

The ALPS study also confirmed the independent influence of mode of delivery 21 .

The ACOG highlights an increased risk of respiratory morbidity in cases of delivery before 39 weeks’ gestation, with significantly higher rates of respiratory failure at 37 weeks’ gestation compared to 39 weeks’ gestation 22 23 .

In terms of a treatment strategy for drug-exposed newborns, this implies that mode of delivery and gestational age should be taken into account in risk stratification. Newborns delivered by caesarean section or pre-term newborns with intrauterine drug exposure could be at increased risk, justifying intensified monitoring. Conversely, a shorter monitoring period may be acceptable for vaginally delivered, full-term newborns without clinical abnormalities.

This clustering of respiratory adaptation disorders observed in our study among newborns with intrauterine SSRI exposure is consistent with the findings of previous studies and with the respiratory distress syndrome (RDS) described in the literature 11 . A systematic analysis by Heli Malm 24 shows that, whilst the literature provides inconsistent findings regarding pre-term birth and small for gestational age (SGA) birth weight, the evidence regarding neonatal adaptation disorders and respiratory symptoms is largely consistent.

The physiological causes remain largely unclear; however, symptoms including ‘respiratory distress syndrome, tremors, restlessness, and increased muscle tone’ 25 26 have been described in newborns exposed to SSRIs. However, it is currently unclear whether these symptoms are indicative of withdrawal symptoms 27 28 or serotonin intoxication 29 30 31 , and this remains the subject of ongoing research.

As there are currently no nationwide guidelines on postnatal care for affected children, there is a considerable need for systematic, national care data. These could form the basis for a standardised postnatal care concept that addresses the specific needs of children with increased neonatal risk. Our study aims to provide a factual basis and offers starting points for developing evidence-based recommendations for action.

Limitations

Despite the findings obtained, our study has several limitations that must be taken into account when interpreting the results.

In addition to the retrospective, single-centre design, the limited number of cases in several subgroups is a key limitation. As a result, effect estimates are associated with wide confidence intervals, and the statistical power for group comparisons is limited.

It was not possible to establish an internal unexposed control group or a hospital-wide comparison cohort comprising all births at the UKER during the study period (2021–2023). Clinically unremarkable newborns who do not meet the criteria for neonatal monitoring as defined in the internal care pathway generally remain in the maternity ward; for these children, comparable data are not recorded with the same level of detail and data structure as for the study population under neonatal monitoring.

The IQTIG data are used in this study for contextual classification. Differences between the groups at University Hospital Erlangen and the IQTIG may be partly attributable to structural differences between the cohorts, the different data collection periods, and differences in detection and coding.

Retrospective data collection does not allow for complete standardisation of clinical event documentation. It cannot therefore be ruled out that the documentation of individual desaturation events was influenced by monitoring intensity, clinical assessment and local documentation practices. This applies in particular to the distinction between short-term, spontaneously reversible monitor events and clinically relevant respiratory events.

Systematic, standardised further diagnostic investigations (e.g. echocardiography, chest imaging) and mandatory aetiological differentiation – such as persistent pulmonary hypertension of the newborn or pneumonia – were not fully documented in the retrospective routine dataset; it was therefore not possible to reliably attribute the respiratory events to specific aetiologies.

Furthermore, the maternal age distribution differs between the University Hospital Erlangen cohort and the nationwide IQTIG comparison group. Whilst the estimated mean age in the IQTIG cohort was 30.4 years, it was 32.4 years in the Erlangen cohort. In particular, the proportion of mothers aged over 35 was significantly higher in Erlangen (35.83% vs. 24.6%). Advanced maternal age may potentially influence neonatal risks and therefore constitutes a confounding factor. One reason for the higher maternal age could be the high proportion of university graduates in Erlangen, as described in the 2022 census (40% compared with 20% across Germany) 32 . As has been described on numerous occasions, a higher level of education correlates with starting a family later in life 33 34 .

The lower pH value observed in our study population warrants critical scrutiny: here, 4.17% of umbilical cord pH readings were below 7.1, whilst this proportion was only 2.22% in the Germany-wide IQTIG comparison. As the mothers in our study population were older at the time of birth than those in the comparison group, it is reasonable to assume that this may have influenced the results as described above. Furthermore, our study cohort showed a higher proportion of post-term (> 41 weeks’ gestation) births, at 7.5%, compared with the IQTIG group (0.5%). The proportion of newborns born between 37 and 41 weeks of gestation was 84.2% in the study cohort, which was lower than in the comparison group (91.0%). These differences must be taken into account as potential factors influencing the lower umbilical cord pH; it is therefore necessary to determine what percentage of the observed pH differences is attributable to intrauterine drug exposure.

Another limiting factor in our study is the lack of documentation regarding the exact dosage of the medications taken and the mothers’ adherence to their medication regimen, meaning that the extent of intrauterine drug exposure cannot be accurately assessed.

Furthermore, there is a possibility of potential double-coding of the ICD-10 diagnoses P22 (‘respiratory distress syndrome in the newborn’) and P28 (‘Other respiratory disorders in the newborn’), which may lead to statistical distortions. The same applies to the categories of the IQTIG group mask ventilation/intubation/oxygen supplementation. It should generally be noted that the assignment of an ICD diagnosis is highly dependent on the given setting, the level of documentation and local billing practices.

Conclusion

In the adjusted analysis, neither gestational age, maternal age nor mode of delivery showed a significant association with the primary end point; nor were there any significant differences between the exposure groups following adjustment. The clinical relevance of the respiratory events observed in the monitored cohort therefore remains unaffected; however, given the retrospective design of the study, a causal interpretation in the sense of a proven drug effect is not possible.

To address the limitations mentioned, future studies should comprise prospective, multi-centre cohort studies with sufficient statistical power for substance-class-specific analyses, as even in large studies the case numbers for individual active substances are frequently too small. It is also necessary to systematically record the severity of the mother’s underlying disease as a key confounder, to establish a standardised definition of poor neonatal adaptation syndrome using validated instruments, and to document the mothers’ dosage and adherence to treatment. To improve comparability, defined reference populations or multi-centre registries with standardised outcome recording should be used; sibling analyses can address familial confounders. National registries could support guidelines on postnatal care and capture both short-term and long-term end points. At the same time, continued development of risk-adapted monitoring strategies remains relevant.

The present findings and the current literature have several specific implications for the daily clinical practice of neonatologists and obstetricians. The S3 Guideline/National Care Guideline on Unipolar Depression (2017) recommends that, in cases of intrauterine exposure to antidepressants, mothers should give birth at a clinic with an attached neonatal unit 14 ; however, the current 2022 edition does not contain any recommendations on postnatal management 13 . Discontinuing or reducing doses during the third trimester is expressly not recommended, as observational data show no reduction in the risk of poor neonatal adaptation syndrome, whilst the risk of maternal relapse increases 35 . Our observations suggest that structured postnatal monitoring can identify clinically relevant events; whether and in what setting this should be carried out requires prospective investigation.

As part of efforts to improve the structure of the healthcare system, rooming-in models should be promoted more actively in order to minimise separation of mother and child and the associated psychological and physical burden. Gynaecologists in private practice should specifically counsel mothers taking antidepressants or anticonvulsants about potential postnatal risks and arrange early referral to suitable perinatal centres. In addition, efforts should be made to strengthen interdisciplinary collaboration between gynaecologists, psychiatrists and neurologists in order to re-evaluate existing medication regimens. Similarly, carrying out close drug monitoring in future could help to ensure the safety of both mother and baby.

International guidelines are converging on a minimum monitoring period of 48 hours; for asymptomatic newborns, monitoring for 48 to 72 hours is recommended, whilst for symptomatic infants, monitoring should continue until complete symptom resolution 36 . Given that, in our cohort, 27.3% of respiratory events did not occur until after 48 hours, our data suggest that the duration of monitoring should be extended on an individualised, risk-adapted basis beyond this minimum standard, particularly in the presence of additional risk factors such as caesarean section, pre-term birth or polypharmacy.

The findings highlight the need for prospective clinical data and may serve as a stimulus for developing standardised local SOPs and, in the longer term, national recommendations. These should ensure a structured approach to caring for affected newborns and enable early risk detection.

Note

Data was collected and analysed by Antonia Hell as part of her doctoral thesis at the Department of Paediatrics and Adolescent Medicine, Friedrich-Alexander University of Erlangen–Nuremberg.