Adequate treatment of mental health disorders during the perinatal period from pregnancy up to one year after birth [1] is essential for maternal and fetal/newborn health [2]. Pharmacological management is recommended for the treatment of moderate to severe perinatal mental health disorders when psychological interventions are not effective [2, 3]. ‘Selective serotonin reuptake inhibitors’ (SSRI) are the most prescribed antidepressant medication in both the general and perinatal populations [4].

Up to 30% of newborns with in-utero SSRI exposure will experience withdrawal signs which are often self-limiting and of short duration [5–7]. These signs include poor muscle tone, tremors, jitteriness, irritability, seizures, feeding difficulties, sleep disturbances, hypoglycemia, and respiratory distress [8–11]. It is unknown if the observed signs are a serotonin syndrome (toxicity) or withdrawal [12, 13], and some authors have described it as a behavioural syndrome [14]. Thus, a range of terms are found in the extant literature with the term withdrawal most commonly used [15–19]. This contrasts with extensive research examining the withdrawal symptoms of newborns from in-utero opioid exposure and a well-defined ‘Neonatal Abstinence Syndrome’ (NAS), now termed ‘Neonatal Opioid Withdrawal Syndrome’ (NOWS) [20].

Few studies of in-utero antidepressant (AD) medication exposure included newborn withdrawal as an outcome measure or considered the effect of newborn feeding method [11, 21–24]. In a small observational cohort study of newborns with in-utero SSRI or other AD medication exposure (n = 247), newborns receiving formula feedings were three times as likely to have withdrawal as compared to newborns who were breastfeeding or receiving mixed feeds [25]. This study, like other studies, included only a small number of newborns who were exposed to antidepressant medications from different categories, leading to difficulty in estimating the effect of in-utero SSRI exposure among breastfed newborns [10, 23]. Therefore, newborn feeding method represents an important area of investigation.

While there is evidence regarding the benefits of breast-feeding among in-utero opioid-exposed newborns on reducing the risk of NOWS, including shortened length of hospital stay, reduced need for pharmacotherapy to treat signs of NOWS, and fewer admissions to NICU [26–28], there is a dearth of research among Neonates With In-utero SSRI Exposure (NeoWISE). Breastfeeding may offer a protective effect to the newborn by transferring low amounts of SSRIs through the breastmilk. While these low amounts have not been associated with adverse outcomes in the newborn [29], this small exposure may be sufficient to mitigate SSRI withdrawal effects. Given there are more pregnant women taking SSRIs than opioids, the study of potential withdrawal following SSRI exposure is warranted. Evidence is needed to confirm whether breastfeeding can reduce the incidence of withdrawal in NeoWISE in order to strengthen existing recommen-dations in perinatal and neonatal care guidelines for this population [7, 8, 13]. Furthermore, it is necessary to determine if NeoWISE would benefit from additional breastfeeding support and follow-up in the first days and weeks after birth.

This study aimed to identify if initiated newborn feeding method was associated with neonatal withdrawal and transfer to the NICU among neonates exposed to SSRIs in-utero. We hypothesized that breastfeeding would be associated with a lower risk of withdrawal and transfer to the NICU due to the benefit of SSRI medication transfer through breast milk and the close skin-to-skin contact with the mother during breastfeeding. This hypothesis was informed by an observational cohort study investigating newborn outcomes following in-utero AD medication exposure [25], and our clinical experience.

A total of 9,926 newborns were included in the NeoWISE Cohort (Supplementary Figure 1). This included 9,014 pregnant women who filled at least one prescription for ≥1 SSRIs during pregnancy. Escitalopram (37%) and sertraline (30%) were the most frequent SSRI medication exposures (Supplementary Table 4). After excluding newborns missing feeding information (n = 2,503), there were 5,079 newborns in the final NeoWISE Cohort with 3,321 (65.4%) exclusively breastfed and 1,758 (34.6%) formula-fed only. Approximately half of the cohort (n = 2,916, 57.4%) had SSRI exposure in trimester three (Supplementary Table 5). The median maternal age (IQR) of exclusively breastfeeding and formula-feeding newborns was 25 years (22–30) and 26 years (22–31) respectively.

As shown in Table 1, there were imbalances in NeoWISE by feeding method. In comparison to breastfed newborns, those who were formula-fed were more likely to have older mothers who lived in an urban area, had a lower neighbourhood income quintile, had a previous live birth, reported smoking at the time of delivery, and gave birth by cesarean delivery. Formula-fed newborns had a slightly lower gestational age. These imbalances in maternal and newborn characteristics were balanced after propensity score weighting (Table 1). Maternal-newborn pairs missing neighbourhood income quintile data (n = 18) were excluded.

The risk of neonatal withdrawal was similar in exclusively breastfed newborns as compared with formula-fed newborns unadjusted (1.5% vs 2.3%, RR = 0.66, 95% CI 0.44, 1.00) and adjusted models (adjRR = 0.86, 95% CI = 0.56, 1.34) (Table 2). In terms of the secondary outcome, there was a lower risk of transfer to the NICU among breastfed newborns as compared with formula-fed newborns in both the unadjusted and adjusted models (7.2% vs 10.8%, adjRR = 0.80, 95% CI = 0.66, 0.97) (Table 2). Based on the E-value, an unmeasured confounder with 1.21-fold association with feeding and NICU admission could account for our observed findings.

In our population-based NeoWISE Cohort, there was an overall low risk of neonatal withdrawal and no difference in risk among exclusively breastfed and formula-fed newborns. This finding persisted in both sensitivity analyses. Our finding regarding the low prevalence of neonatal withdrawal contrasts with the results of higher rates in other studies, ranging from 11.2% to 63.2% [11, 25, 39, 54–57]. The vast differences in withdrawal prevalence may be attributed to the outcome variable definition, the method used to identify withdrawal signs in newborns with in-utero AD exposure, and the selected cut-off scores. In a recent population-based cohort study in the USA, Cornet et al. reported that 11.2% of newborns with exposure to SSRIs after 20 weeks gestation had delayed neonatal adaptation; delayed neonatal adaptation was defined as a 5-mins Apgar score ≤5, resuscitation at birth, or admission to the NICU for respiratory support. Levinson-Castiel et al. reported neonatal abstinence syndrome occurring in 30% of newborns, 13% of whom had severe symptoms, in their cohort study of newborns exposed to SSRIs or venlafaxine [55]. Neonatal abstinence syndrome was identified by a Finnegan score of ≥8 as severe, 4–7 as mild and 0–3 as normal [55]. Oberlander et al. [39] and Chambers et al. [11] reported similar rates of approximately 30%.

Ferreria et al. reported that 63.2% had one or more neonatal behavioural signs in their retrospective chart study of newborns with in-utero SSRI or venlafaxine exposure (n = 79) [56]. Kieviet et al. reported a rate of 64% of newborns diagnosed with withdrawal and 20% of these had severe symptoms in their observational cohort study of newborns (n = 247) in the Netherlands with at least third-trimester in-utero exposure to serotonergic AD exposure (i.e. SSRI, SNRI, or NaSSA), [25]. Withdrawal was defined as at least one Finnegan score of ≥4 during the inpatient hospitalization, with ≥4 being mild and ≥8 being severe withdrawal [25].

It is not possible to know how clinicians assessed the newborns in our study; however, the threshold to identify withdrawal may have been higher than prior studies given our lower rate of withdrawal. Clinicians may be using clinical practice guidelines that were intended for caring for newborns with opioid exposure to inform their clinical care and management of newborns with other in-utero substance exposures, given that no validated instrument exists to assess for the presence of neonatal SSRI withdrawal. For example, according to an Ontario guideline, if newborns were assessed using the ‘Finnegan Neonatal Abstinence Scoring System’ and there were three consecutive scores ≥8 or a score of ≥12 (either an average of 2 scores or two consecutive scores of ≥12) [58], they may have received a diagnosis of withdrawal and thus be assigned the P96.1 diagnosis code. This higher threshold may have contributed to our study’s observed low rate of withdrawal.

In a secondary analysis of two observational studies (n = 214), Yang et al. [59] reported that 33% of newborns had withdrawal signs which were significantly higher in preterm as compared to full-term newborns (54% vs 31%, p = 0.020) [59]. Only 1.4% of newborns in their study were noted to have a Finnegan score of ≥8 indicating the need for intervention, with restless sleep, tremors and mottling as the most common withdrawal signs reported [59]. Their findings of a 1.4% rate of withdrawal signs [59] is more consistent with our study results (1.5% and 2.3% for exclusively breastfed and formula-fed newborns, respectively). Our study included both term and preterm newborns.

The setting may also have influenced study results. Participants recruited by Kieviet et al. [25] were from a psychiatric, obstetric and paediatric clinic that included providers with psychiatric disorder expertise. It is possible that these participants had more severe perinatal mental health disorders as compared to our population-based sample where mothers were beneficiaries of a publicly funded drug benefit program. Study participants receiving care at the specialized centre described by Kieviet et al. [25] may have had more focused assessment and monitoring during their postnatal inpatient stay than our study participants. In Ontario, there are few integrated perinatal mental health programs and the existing programs are predominantly provided on an outpatient basis [60]. Therefore, we can assume that most newborns in our study likely received the same routine monitoring and care as newborns unexposed to SSRIs.

During our study timeframe, the Ontario Government provided concerted efforts to implement the ‘Baby Friendly Initiative’(BFI) [61–63]. The BFI recommendations include providing skin-to-skin contact for all newborns regardless of feeding method, promoting and encouraging maternal-newborn togetherness, and initiating early and frequent breastfeeding, to name a few. Efforts to implement these evidence-based practices in Ontario hospitals may have contributed to sustaining the higher rate of exclusive breastfeeding (65.3%) observed in our study compared to the rates reported by Kieviet et al. [25] (43.3% (n = 107) breastfeeding, 18.2% (n = 45) formula feeding and 38.5% (n = 95) mixed feeding).

We hypothesized that breastfeeding might confer a significant risk reduction for newborn withdrawal given that Kieviet et al., had reported that formula-feeding newborns in their study were three times as likely to have withdrawal as compared to newborns who were breastfed or received mixed feeds (OR = 3.16, 95% CI 1.40, 7.13) [25]. The overall low rate of withdrawal signs observed in our study may be due to newborns receiving non-pharmacological support (e.g. skin-to-skin contact, rooming-in, breastfeeding), which has been reported to be effective in reducing the severity of withdrawal among opioid exposed newborns [26–28]. Formula-fed newborns in our study likely received benefits from some of these non-pharmacologic interventions, which could have partially affected our study result of no difference in risk of withdrawal by feeding method. Additionally, we defined our cohort as exposure to any SSRI, while Kieviet et al. [25, 64] included multiple medication categories (i.e. SSRIs, SNRIs, NaSSA).

Serotonergic antidepressants have different pharmacokinetic properties, and thus, they are likely to impact the exposed newborn differently after birth. Cornet et al. [54] suggested that there may be a causal relationship between type and dose of SSRI medication exposure and risk of delayed neonatal adaptation, although research is required to identify the mechanism underlying this relationship [54]. These authors did not include feeding method as a variable in their analysis; however, they acknowledged that the effect of sudden discontinuation of SSRI exposure after birth as compared with continuous exposure via breast milk was an area of investigation they did not evaluate [54].

Strengths of our study include the large sample size of NeoWISE compared to other cohorts; the use of a population-based cohort study design that linked multiple health administrative databases and validated registries; and IPTW of propensity scores to balance differences in baseline characteristics related to newborn feeding methods minimizing confounding bias [46].

This study has some limitations. The cohort was restricted to women on ODB as this was the only record of SSRI prescription claims in ICES. The results are likely not generalizable given that the study cohort represents less than 1% of the population of births in Ontario within the study timeframe, and thus, different effects of breastfeeding may be observed in the larger pregnancy population. The NeoWISE cohort was much younger than the general population of childbearing women in Ontario (Supplementary Table 6). Women with a lower income, a younger age, and poor mental health are at increased risk of poor breastfeeding outcomes [70]. Additional maternal and newborn characteristics associated with low rates of breastfeeding include primiparous, unmarried, low social support, maternal obesity, diabetes during pregnancy, smoking during pregnancy, cesarean birth, preterm infant, admission to the NICU, and formula supplementation in hospital [70]. Regardless of these potential barriers, we observed that the majority of mothers in our cohort of ODB recipients were either exclusively breastfeeding or mixed feeding at the time of postnatal hospital discharge.

We did not collect data on maternal mental health diagnoses or the severity of mental health concerns. SSRI medications are primarily prescribed to treat depression, as observed in other retrospective studies [56]. However, SSRIs have also been used as a first-line pharmacotherapy for many other psychiatric disorders. It is possible that NeoWISE who were exposed to SSRIs through the entire pregnancy (n = 1,636, 32.2%) had mothers with more severe mental health concerns, and this may have influenced observed associations. Off label SSRI prescribing has been reported for binge eating disorders, body dysmorphic disorder, fibromyalgia, paraphilias, autism, and Raynaud phenomenon [71], although these are likely rare occurrences in the perinatal population.

While there were no guideline changes in medication prescribing or treatment for perinatal mental health disorders during the study timeframe, the prescribing practices of health care providers may have evolved as more information became available on the risks/benefits of various psychotropics. By using medication dispensing data, we reduced bias due to confounding by indication by selecting only pregnant women who were dispensed an SSRI medication, and we were able to conduct sensitivity analysis comparing any SSRI exposure during pregnancy versus third trimester SSRI exposure. Although we were able to identify the trimester in which the SSRI medication was dispensed by using data obtained from the ODB, it is not possible to know for certain if the mother took the medication as prescribed. Therefore, medication exposure may have been misclassification when assembling the study cohort.

Data were limited to variables that were available in health administrative databases. The specific withdrawal signs of the newborns in our study are unknown, given that a diagnostic code (P96.1) was used as the primary outcome. As a result, we have likely underestimated the incidence of withdrawal given that the P96.1 variable may have reduced sensitivity to identify newborns who are experiencing minor adverse reactions following in-utero SSRI exposure. Numerous authors have reported that some of the most common signs of withdrawal include tremors, irritability, crying, sleep disturbances, increased moro reflex [13, 14, 56, 72–74] and these are not variables in health administrative databases. Furthermore, the P96.1 diagnosis code is non-specific to SSRI exposure, so all clinicians may not use it. Currently, no variable exists in health administrative databases for SSRI exposure.

To date, past research has many limitations, including the predominant use of case review methodology, small sample sizes, self-reported adverse outcomes, and a lack of long-term follow-up [75]. Pharmacoepidemiologic studies require more robust breastfeeding data. Jordan et al. [76] emphasized the need for quality breastfeeding data in population health databases to enable pharmacoepidemiologists to explore relationships between medication exposures, breastfeeding, and short and long-term newborn outcomes. Health care providers can strive to ensure complete and accurate documentation of newborn feeding; audit and feedback cycles may be useful to demonstrate progress in this area. Additional methods of capturing newborn feeding data are warranted. A prospective cohort study design may facilitate improved data collection on essential variables such as mental health status, medication history, newborn feeding details, and neonatal behavioural/withdrawal signs. More rigorous data collection would improve surveillance and reporting of adverse outcomes, mainly to obtain a more precise estimate of the withdrawal prevalence in the NeoWISE.

Future researchers should consider creating a core outcome set of variables reflecting the most common signs of SSRI withdrawal, as has been done for in-utero opioid exposure [77]. Wang and Cosci have proposed a framework that is worth consideration [78]; transient neonatal behavioural signs, which may impact newborn feeding ability, will need to be integrated. This promising research direction may improve the sensitivity and specificity of identifying newborns with clinically relevant withdrawal. Finally, capturing parent-reported observations in future research studies would be particularly important. Qualitative studies exploring the parental experience of caring for NeoWISE may provide insightful guidance for clinical care, particularly regarding the pre- and postnatal support women require to achieve their desired newborn feeding goals.

In this study, we report an overall low risk of neonatal withdrawal, and no difference in risk among breastfed compared with formula-fed NeoWISE. We also observed a possible reduced risk of transfer to the NICU in breastfed newborns. The findings of this study contribute to the extant literature, which indicates that perinatal women who take SSRI medications can be supported in their preferred newborn feeding method with the appropriate monitoring for newborn well-being. Pregnant and breastfeeding families need to receive anticipatory guidance about potential signs of withdrawal in the NeoWISE and when to seek health care after birth.