Impact of maternal medication for opioid use disorder on neurodevelopmental outcomes of infants treated for neonatal opioid withdrawal syndrome

2.1 Study design and participants

As reported in our previous work [12], we retrospectively extracted data from mother-infant dyads’ hospitalization. Maternal and infant/child data were obtained from the hospital based electronic medical records between July 2006 and December 2018 from a single tertiary care academic children’s hospital in Massachusetts, with approximately 4000 births per year. De-identified infant hospital data was then linked to data obtained from the Massachusetts Department of Public Health (DPH) Early Intervention database. The infants were either inborn or, had been transferred from a regional Level 1 nursery, between July 2006 to June 2013, due to either non-NOWS related diagnosis requiring critical care, difficulty capturing NOWS despite maximizing treatment or, a complication secondary to NOWS such as seizures. During this time period, infants with clinically significant NOWS, defined as need for pharmacological treatment, were managed in the neonatal intensive care unit (NICU). The NICU during the study period used combination therapy for treatment of NOWS with neonatal morphine sulfate as the primary agent with either phenobarbital or clonidine as an adjunct medication. Efforts were also made to promote non-pharmacologic care by promoting skin to skin care, breastfeeding as eligible and providing quiet environment as much as feasible in a NICU setting of care provision. The hospital policy was to refer all infants with the NOWS diagnosis cared for in the NICU to EI services, even before NOWS became an automatic qualifying diagnosis for EI services beginning in 2009. Infants in this cohort were initially identified using ICD-9 diagnostic codes for “maternal drugs affecting newborns” or “neonatal abstinence syndrome,” two specific codes for substance-exposed infants. The maternal and infant electronic medical records (EMR) were then reviewed for all infants with these ICD-9 codes and, only those with confirmed in-utero opioid exposure from maternal history, positive urine toxicology for mother and positive urine/meconium toxicology for infants were included in the final cohort. Infants with only non-opioid in-utero exposures with potential for neonatal withdrawal symptoms were excluded from the cohort. Additionally, infants with chromosomal anomalies and, those born < 35 weeks gestational age, identified in the medical records, were excluded from the study because they would be at high risk for delays in neurodevelopment irrespective of the NOWS diagnosis. This study was approved by Baystate Health Institutional Review Board and the Massachusetts DPH Institutional Review Board.

2.2 Descriptor variables

The following variables collected at the time of hospital discharge from the EMR were used to describe the cohort of infants with NOWS and, their mothers and, to assess associations with receipt of EI enrollment and services. Infant characteristics included: race or ethnicity (categorized: Non-Hispanic White, Hispanic/Latino, Black, other, unknown), sex (male/female), birth weight (kilograms), gestational age (weeks), any period of hospitalization in the NICU (yes/no), length of hospital stay (weeks), any breastfeeding during the hospitalization (yes/no), and NOWS medication type [none, neonatal morphine sulfate (0.4 mg/ml oral solution) (yes/no), phenobarbital (20 mg/5mL oral solution) (yes/no), and clonidine (10 mcg/mL oral solution) (yes/no)]. Maternal characteristics included age at time of infant’s birth (years), psychiatric diagnoses (bipolar disorder, depression, anxiety, none), any psychiatric medications (yes/no), tobacco use in pregnancy (self-report yes/no), marijuana use in pregnancy (self-report or laboratory result, yes/no), type of MOUD medication (methadone, suboxone/buprenorphine), and duration of time in MOUD (months) for those enrolled. We were not able to collect data pertaining to maternal education level, socioeconomic status, eligibility/receipt of public assistance and social support for this study. These descriptor variables were selected based on the published literature about NOWS risk and, severity, clinical experiences, and, our above described hypotheses. Additionally 4 cohort subgroups were identified - infants whose mother were already enrolled in MOUD before pregnancy was diagnosed (Group 1), infants whose mother enrolled in MOUD after pregnancy was diagnosed (Group 2), infants whose mother were actively using illicit drugs and not in MOUD (Group 3) and infants whose mother were on prescribed opioid for chronic pain (Group 4).

2.3 Outcome variables

EI enrollment (categorized as enrolled at 6, 12, and 24 months of age) and duration of EI services provision within the first two years of life were our primary outcome variables of interest. This matched data was obtained through Massachusetts DPH EI database. Additionally, during the study period the child’s electronic medical record was reviewed for documentation of BDD diagnoses, as determined during care provision visits by a primary care pediatrician, pediatric neurologist and/or, development behavioral specialist. The age of the subjects at the time of BDD diagnoses ranged from approximately from 5–12 years. These outcome variables of interest included attention deficit hyperactivity disorder (ADHD), oppositional defiant disorder (ODD), obsessive compulsive disorder (OCD), child anxiety, child depression, mood disorder not otherwise specified (NOS), cognitive delay, speech delay, motor delay and, other.

2.4 Statistical analyses

Descriptive statistics were calculated for the infant and maternal variables of the entire cohort as well as for the subset of infants that were enrolled in EI. Continuous variables were described using means and, standard deviation and compared using t-tests. Categorical variable were described using frequencies and percents, and compared with Fisher’s exact test.

Statistical analyses were performed first, focusing on comparing MOUD initiation before pregnancy was diagnosed (Group 1) and, MOUD initiation after pregnancy was diagnosed (Group 2). Our initial intent was to compare neurodevelopmental outcomes between those exposed to MOUD medications in-utero with those with other opioid exposure either non-prescribed or prescribed for chronic pain issues. However, since most mothers were initiated on MOUD even if in the last trimester the number of subjects in Groups 3 and 4 were too small to make a meaningful comparison and were excluded from the analyses

We performed separate logistic regression analyses for the two primary outcomes (any behavior and/or, developmental disorder diagnosis and EI service at 2 years) to determine whether variables were independently associated with each outcome. Since the outcome –any BDD –had the larger sample size (n = 196), we used backward selection for this outcome in which all variables having a critical p-value < 0.25 were retained for consideration. MOUD history and, gestational age were included in the final model regardless of their significance. The variables that remained in the model for any behavioral diagnosis were also used for the outcome EI services at 2 years (n = 58).

3.1 Overall cohort characteristics (Table 1)

In our cohort of 242 infants, 96 (39.7%) OENs were born to mothers enrolled in MOUD before pregnancy was diagnosed (Group 1); 102 (42.2%) were born to mothers who enrolled in MOUD after pregnancy was diagnosed (Group 2), 44 (18.2%) were born to mothers who were actively using illicit opioids at time of birth and not enrolled in MOUD and/or were born to mothers who were on prescribed opioids for a diagnosis of chronic pain disorder (Group 3 and 4). Of these 242, we had BDD data available for all 242 infants (age range between 5 years to 12 years at the time of follow-up), in our EMR database. Of these 242 infants 32.6% had one or more BDD diagnosis. At the time of discharge, 184 infants (76.0%) were referred for EI services of which 113 (46.7%) were further evaluated and, enrolled for EI services.

For the overall cohort of OENs the infant characteristics were notable for a mean gestational age of 38.6 weeks (SD 1.7) weeks at birth, with mean birth weight of 3.0 Kg (SD 0.5). Majority of the infants were of non-Hispanic White race/ethnicity (71.5%), male (53.3) with an average hospital stay of 23.7 days (SD 11.5). Twenty percent of these infants received any maternal breastmilk or, breastfeeding during their hospitalization. Of the infants who needed treatment, morphine was the first line medication (96.7%) with adjunctive therapy as phenobarbital or clonidine per the NICU protocol. One quarter of infants included in the study were discharged from the hospital into foster care. Maternal data exhibited that average age at the time of delivery was 28.2 (SD 5.2) years; mothers had an additional psychiatric diagnoses including 15.3% with bipolar disorder, 40.5% with depression and 26.0% with anxiety and 33.5% were on one or more medications for these diagnoses in addition to MOUD. Of the mothers enrolled in MOUD, 66% were on methadone and 34% were on buprenorphine with 35.8% using illicit drugs during pregnancy. Additionally, more than half (57.4%) of the mothers were using tobacco during pregnancy with another 16.9% reporting marijuana use during the pregnancy. Similar results were found for infants who had been enrolled in EI and, had matched EI data available.

We had hypothesized that infants born to mothers who were in stable MOUD programs prior to diagnosis of the current pregnancy (Group 1), a marker of more stable recovery process, would have similar outcomes as compared to mothers who enrolled in MOUD program after pregnancy was diagnosed (Group 2). For this purpose we compared the maternal and, infant birth hospitalization characteristics for these two groups and instead found that infants born to mothers in Group 1 were born at higher gestational age (GA) of 39.0 (±1.5) vs 38.4 (±1.8) weeks (p = 0.01); had a higher birth weight (BW) 3.1±0.5 vs 2.9±0.5 Kg (p = 0.02) and less likely to be discharged to foster care 8.3% vs 34.3 % (p = < 0.001). There were no differences in sex, ethnicity, any mother’s breast milk feeding or breastfeeding, need for pharmacotherapy for NOWS, length of hospital stay, any behavioral diagnoses, EI enrollment and/or, duration of EI services. However, in Group 1 we found lower rates of additional in-utero exposures for tobacco 51% vs 66.7% (p = 0.03), marijuana 10.4% vs 22.5% (p = 0.02) as well as lower rates of maternal illicit drug use while enrolled in MOUD 8.3% vs 39.2% (p = < 0.001).

Among infants with additional follow-up visits in our EMR post discharge, 32.6% OENs had one or, more BDD diagnosis. We compared the maternal and, infant characteristics among OENs with and without BDD diagnosis to identify any associated risk exposures. The infants with a BDD diagnosis were born at similar GA 38.4 (±1.7) vs 38.7 (±1.7) (p = 0.33), birth weight 2.9 (±0.5) vs 3.1(±0.5) as well as similar rate of neonatal morphine sulfate administration as first line of treatment for NOWS 96.2% vs 96.9% (p = 0.72) but slightly higher rates of phenobarbital as an adjunct 79.7% vs 68.1% (p = 0.07) and, lower rates of clonidine as adjunct 11.4% vs 28.8% (p = 0.003). We found a significant difference by race and BDD status (p < 0.001). Infants with Hispanic/Latino and Other/unknown race/ethnicities had somewhat lower rates of BDD as compared to infants of white or black race/ethnicity. With respect to maternal characteristics, mothers of infants with a BDD diagnosis were slightly more likely to be treated with methadone for MOUD 75% vs 61.2 % as opposed to buprenorphine 25% vs 38.8% (p = 0.08). However there were no differences in maternal age at time of delivery, maternal psychiatric diagnoses (bipolar disorder, depression, anxiety), need for medication for the aforementioned psychiatric diagnosis, additional exposures to alcohol, tobacco and/or marijuana and illicit drug use while in MOUD.

There was no difference in the prevalence of BDD diagnoses for infants born to mothers who initiated MOUD pre or, post-pregnancy diagnosis.

With respect to our primary outcome of interest, enrollment in EI services at two years of age, a smaller model emerged due to the smaller sample size (n = 58). In this model no variable was identified as significant for the primary outcome.

For the secondary outcome, any BDD diagnosis, neither timing of maternal MOUD enrollment nor gestational age was associated. However, NOWS treatment with clonidine was associated with lower odds of these diagnoses by 62% (OR = 0.38, 95% CI: 0.16, 0.91). Interestingly a reported anxiety disorder for the mother more than doubled the odds of any BDD diagnosis (OR = 2.15, 95% CI: 1.07, 4.34).

4.1 Early neonatal outcomes

OENs born to mothers, who were actively using illicit opioid drugs before pregnancy was diagnosed with subsequent enrollment in MOUD once pregnancy was diagnosed, were born at lower gestation and had lower birth weights, as compared to mothers who were enrolled in MOUD programs before diagnosis of pregnancy. This is in agreement with previously reported data [2, 3], and, is probably secondary to late entry to prenatal care, as well as higher rates of additional exposures as alcohol, tobacco and, marijuana and continued illicit drug use even after maternal enrollment in MOUD programs [16–18]. This was unrelated to the maternal-infant ethnicity which were similar in both the groups, in this primarily non-Hispanic White cohort. We found similar breastfeeding rates in both the groups which while low in this historic cohort as compared to more recent data are still comparable to reported rates in the US [19, 20]. Since our cohort comprised of infants with severe NOWS symptoms requiring pharmacotherapy it was not surprising to note the similarly high utilization of the medications for both groups. Due to concerns of negative impact of phenobarbital on a developing brain [21, 22], our center conducted a randomized clinical trial comparing phenobarbital with clonidine as an adjunct to neonatal morphine sulfate as the primary agent for pharmacotherapy for NOWS [23]. This led to a change in our center’s clinical practice guidelines in 2012 to primarily use clonidine as the adjunctive medication while limiting phenobarbital use to specific cases, which explains the three medications in our cohort.

In our cohort, we found that a greater percent of OENs were discharged home to foster care guardianships in Group 2 (MOUD initiation after pregnancy was diagnosed). MOUD initiation before pregnancy diagnosis may be associated with greater social support, health literacy/education around OUD, pregnancy and NOWS, which could potentially influence custody placements. Our data suggest that MOUD initiation before pregnancy was diagnosed was associated with a more stable recovery as noted by lower use of illicit drugs while enrolled in MOUD as well as significantly lower additional exposures as tobacco and/or marijuana in pregnancy. Active drug use at the time of birth raises concerns of child abuse or neglect as well as caregivers ability to provide care to the newborn and thus can lead to removal of custody from the biological parent [24]. The short term neonatal outcomes from our cohort are in concordance with the other published data, wherein maternal stable rehabilitation improves pregnancy outcomes, strongly endorsing and supporting recovery programs for women in child bearing age with a diagnosis of OUD.

4.2 Long term behavioral and neurodevelopment outcomes

Almost a third of infants in our cohort were diagnosed with a behavioral and/or developmental disorder, a significantly higher number than reported for the overall US population (17.4% for children age 2–8 years) [25, 26]. While we acknowledge that our specific cohort is at the highest risk for these diagnoses due to the severity of NOWS, the numbers are much higher than anticipated. Interestingly for our cohort we noted a significant protective effect of Hispanic/Latino and other ethnicity on occurrence of BDD, this is in contrast to previously published work [27]. We have no clear reasoning for this and suggest there is a need for additional research that examines differences by race/ethnicity. A slightly higher phenobarbital and lower clonidine use as an adjunct treatment of NOWS and maternal prenatal diagnosis of anxiety disorder correlated with an infant diagnosis of BDD. The rates of near late preterm or near term infants have consistently been on the rise over the last decade and are known to be associated with poorer neurodevelopmental outcomes in the general population [28]. It is highly possible that the higher rates of BDD diagnosis in our cohort may be by chance, as this association with gestational age did not persist in the regression model. Both morphine [29] and, phenobarbital [21, 22], in animal and human studies have shown to impact structural brain development as well as poorer functional outcomes. In addition, as reported by Rees et al., maternal depression and anxiety are the most common mental health problems during pregnancy, with approximately 12% of women experiencing depression and 13% experiencing anxiety, with many experiencing both [30]. Other studies have noted similar effects of maternal mental health diagnoses on the outcomes for their children, especially with respect to prenatal anxiety [31]. Thus unsurprisingly we found that both medications for NOWS treatment and maternal anxiety disorder diagnosis increased the risks of BDD in our cohort of OENs. Additionally, maternal prenatal diagnosis of anxiety disorder was the only predictor with a trend for EI service utilization at 2 years of age, suggesting that greater attention should be given to maternal mental health history during the prenatal care.

Both maternal MOUD with buprenorphine and infant NOWS treatment with clonidine seem to confer a protective effect for the diagnosis of BDD, though the effect size was not statistically significant for buprenorphine after adjustment. In animal models both methadone and buprenorphine have shown alterations in the myelin sheath formation with in-utero exposures [32, 33]. Also, multiple studies have noted better short term neonatal outcomes with buprenorphine as compared to methadone with less severe NOWS and thus it is considered as the preferential choice for maternal MOUD [34]. Clonidine is a classical α-2 adrenergic agonist with known neuroprotective abilities. OENs have multiple exposures that can impair structural brain development, both in pre- as well as postnatal period, hence it is heartening to see that a treatment modality can ameliorate some of the impact from these noxious exposures.

As hypothesized, we found that the timing of maternal MOUD enrollment did not appear to have any association with the rates for BDD and/or receipt of EI services at two years of age. The general assumption is that the stability of maternal recovery status would improve these outcomes. Despite higher gestational age and birth weight, lower exposure to illicit drugs while in MOUD, tobacco and marijuana and discharge home with the biological families the infants continued to be at significant risk for these negative outcomes. These findings suggest that chronic daily in-utero exposure to MOUD opioid medications may have a significant hitherto undemonstrated long-term effect and may require further study among non-opioid based MOUD protocols, for example treatment with the opioid antagonist Naltrexone. Additionally, these risks must be considered to increase access and adherence to EI services However, the cognitive and behavioral outcomes association with MOUD exposure, is likely highly complex [35, 36] and future studies that are able to control for the multitude of factors known to impact a child’s growth and development are urgently needed.

4.3 Strengths and limitations

The major strength of our study, the ability to correlate a single center patient level data with long term outcomes, may also limit our ability to make our data generalizable to other populations. However, since the nature of the measured clinical and outcome variables are not subject to a lot of variation over time, we believe that our data would be highly replicable if similar analyses are done by others. As EI participation is a voluntary choice by the parents we may not have a completely accurate representation of our entire cohort, however, being the only children’s hospital for entire Western MA a majority of infants in this cohort post-discharge will follow-up at our subspecialty clinics for any identified developmental delays allowing us to have robust data about BDD. As stated above, the inability to control longitudinally for psychosocial and socioenvironmental factors that may contribute to BDD, including maternal health and recovery, is a limitation that will need to be addressed by future studies.