Neonatal Opioid Withdrawal Syndrome: A Comparison of As-Needed Pharmacotherapy Free

We performed a retrospective cohort review of infants pharmacologically treated for NOWS at 4 Massachusetts hospitals between 2017 and 2021. This project was a secondary analysis of data collected by the Massachusetts Perinatal Quality Improvement Network (PNQIN). PNQIN is a perinatal quality collaborative that includes birthing hospitals as well as the Massachusetts Department of Public Health, Department of Children and Families, and Health Policy Commission.¹⁷  Participation in PNQIN is voluntary. Since 2017, PNQIN has coordinated a statewide quality improvement initiative focused on improving care for infants affected by NOWS. Hospital teams choosing to participate in this initiative abstract and enter patient-level data into a statewide deidentified REDCap database, including information on maternal demographics, maternal drug exposures, birth demographics, NOWS care protocols, NOWS pharmacologic treatment, and hospital readmissions. Individual hospital teams are multidisciplinary and include physicians, nurses, social workers, lactation specialists, and others.

Thirty-seven level I, II, and III birthing hospitals are participating in the PNQIN NOWS initiative. At time of delivery, 83% of pregnant women with opioid use disorder in the network are receiving medication for opioid use disorder (MOUD).^17,18  The majority of participating hospitals (including our 4 hospitals) follow institution-specific guidelines for the management of pregnant people with opioid use disorder and infants at risk for NOWS. The statewide quality improvement project was deemed nonhuman subjects research by the PNQIN leadership team hospitals institutional review boards, with all participating sites having a data use agreement with PNQIN for data sharing. For this study, we examined the 4 hospitals that had transitioned from scheduled opioid dosing to PRN dosing before the study time period with complete data in the database. The 4 sites had transitioned from scheduled to PRN dosing as part of the implementation of eat, sleep, console (ESC) assessment and management protocols. All sites underwent standardized training workshops organized by PNQIN and adapted the formal ESC NOWS Care Tool (2nd Edition), with the associated educational manuals, training cases, and reliability tools.^19,20 

Three of the 4 study centers administered morphine as the primary PRN pharmacologic agent and 1 center administered methadone as the primary PRN pharmacologic agent. Pharmacologic treatment protocols are included in Supplemental Table 4. All hospitals had enthusiastically adopted and ensured extensive staff education about the ESC tool before the study time period and had defined parameters for when infants would transition from PRN dosing to scheduled opioid dosing (Supplemental Table 4). Infants receiving frequent PRN dosing (range 3–6 doses in a specified time frame, usually within 24–48 hours) with continued symptoms of withdrawal were placed on scheduled opioids to adequately control symptoms and subsequently required an opioid taper to avoid iatrogenic withdrawal. All 4 hospitals practiced a rooming-in model of care in which infants were able to stay with their parent for the duration of infant hospitalization, even after maternal hospital discharge, with a nonpharmacologic care bundle in place. Specifics of the nonpharmacologic care bundles used at the 4 centers are listed in Supplemental Table 4. All infants are monitored in-hospital for at least 24 to 48 hours after the final dose of opioid medication for NOWS treatment before being discharged.

Infants born ≥36 weeks gestation with prenatal opioid exposure (nonprescribed or prescribed MOUD) and requiring pharmacologic therapy with initial receipt of PRN dosing were included. Infants born to mothers receiving naltrexone during their pregnancy were excluded because this medication is not associated with NOWS. All data were reviewed to ensure completeness, and site leaders were contacted regarding incomplete data before final analyses.

Maternal variables examined include birth hospital, race/ethnicity, age at delivery, MOUD exposure, and use of any nonprescribed drugs, nicotine, selective serotonin reuptake inhibitors (SSRIs), and/or benzodiazepines. Nonprescribed drug use was defined as any unprescribed use of opioid medications or substances of abuse during the pregnancy, excluding nonprescribed use that only occurred in the first trimester, as documented in the medical record by patient history or toxicology screenings. Psychiatric medication and nicotine exposure were defined as any use during the pregnancy per medical record review. Infant variables examined include infant sex, gestational age at delivery, birth weight, infant care location, eligibility to receive mother’s breastmilk, receipt of mother’s breastmilk in hospital and at discharge, discharge from the hospital with a biological parent, and any 30-day readmission or emergency department visit. Breastfeeding eligibility criteria were hospital-specific but, in general, relied on a commitment to an addiction recovery program, prenatal care, and negative toxicology screening for unprescribed substances around the time of delivery.

Our primary outcome measures include the percentage of infants who transitioned to scheduled opioid medication dosing, hospital LOS, and the number of PRN doses administered to infants treated with PRN methadone versus PRN morphine. Additional measures and maternal covariates included maternal MOUD exposure, infant care location, any breastfeeding in hospital and at discharge, infant sex, and maternal exposure to nonprescribed substances, nicotine, SSRIs, and/or benzodiazepines.

Mixed effects logistic and linear regression models were employed to evaluate differences in transition rate to scheduled opioid dosing, LOS, and the number of PRN doses administered to infants treated with PRN methadone versus PRN morphine. Infants treated with morphine were used as the reference group. We examined covariates for association with key NOWS outcome measures and examined differences in these covariates between treatment groups. Adjusted and unadjusted odds ratios or risk ratios with 95% confidence intervals (CI) were reported in the final model, taking into account key covariates that met statistical significance. We used multivariable mixed effects linear regression models and adjusted for fixed effects covariates and adjusted for hospital site to estimate site-specific effects. In unadjusted analyses, models accounted for hospital-level random effects. All analyses were performed by using SAS statistical software (version 9.4; SAS Institute, Inc).

Of 138 total infants, 86 infants were treated with PRN methadone and 52 were treated with PRN morphine. During the study period, 38.4% of infants received pharmacologic treatment. There were no significant differences between methadone and morphine groups in maternal race/ethnicity or maternal age at delivery, with a majority of mothers identifying as non-Hispanic White (Table 1). No significant differences between mothers of infants in the methadone versus morphine groups were noted regarding maternal nonprescribed drug use, nicotine use, SSRI use, or benzodiazepine use (Table 1). There were no significant differences in infant demographics between the methadone and morphine groups (Table 1). There was a significant difference in the number of infants treated with methadone who received care on the pediatrics unit compared with those receiving morphine because of differences in institutional NOWS protocols (Table 1). No significant differences in eligibility to receive mother’s breastmilk or infant’s receipt of mother’s breastmilk in hospital and at time of discharge between methadone and morphine groups were identified. There were also no significant differences in infants discharged from the hospital with a biological parent or 30-day readmission/emergency department visit rate between groups.

There were differences by center in pharmacologic treatment rates (34.0% at Center A, 39.3% at Center B, 25.8% at Center C, and 43.7% at Center D; P = .03); however, when comparing centers that used methadone versus morphine, the difference was not statistically significant (43.7% versus 33.0%, P = .05). In assessing for differences in nonpharmacologic care variables, there were no significant differences in rooming-in after maternal discharge among the 4 centers; however, some differences were noted in rooming-in rates before maternal discharge and breastfeeding initiation rates (Supplemental Table 4). However, when comparing the morphine centers (A, B, and C) to the methadone center (D), there were no statistically significant differences in rooming-in rates before (73.1% vs 82.5%, P = .08) or after maternal discharge (63.1% vs 61.7%, P = .85).

There was no significant difference in mean LOS between groups in adjusted and unadjusted models (adjusted for maternal MOUD, NICU utilization, breastfeeding, nonprescribed drug exposure, benzodiazepine exposure, hospital random effects), respectively (adjusted: risk ratio [RR] 1.12, 95% CI 0.80–1.43) (Table 2). LOS was significantly shorter in the infants treated with PRN opioids alone (mean 10.4 days [standard deviation (SD) 4.7] for morphine and mean 11.3 days [SD 4.4] for methadone) compared with those who transitioned to scheduled opioids (mean 24.5 days [SD 7.9] for morphine and mean 28.8 days [SD 12.3] for morphine). No significant differences were observed in the average number of PRN doses received in the methadone versus the morphine groups (adjusted: RR 0.65, 95% CI 0.41–1.02). Lastly, no significant differences in rates of transition to scheduled dosing were observed between the methadone versus the morphine groups (adjusted: odds ratio [OR] 1.02, 95% CI 0.88–1.19) (Table 2).

Several covariates were associated with significant differences in hospital LOS, the number of PRN doses received, and rates of conversion to scheduled dosing (Table 3). There were significant differences in NOW outcomes when mothers had received MOUD with methadone, infants received care within the NICU, and when breastfeeding occurred during the hospitalization and at discharge (Table 3). Lastly, the analysis of additional maternal exposures on key NOWS outcomes revealed differences in all 3 NOWS outcomes in the setting of maternal nonprescribed drug use and maternal benzodiazepine exposure during pregnancy (Table 3).

In a small, retrospective cohort across multiple sites, we found no significant differences in key NOWS hospital outcomes in infants treated with methadone versus morphine on a PRN basis. Infants in both the methadone and morphine groups had similar LOS, the number of PRN doses received, and rates of conversion to scheduled opioid dosing. However, the effect of several important covariates such as breastfeeding, infant care location, and maternal factors like nonprescribed substance or benzodiazepine exposure, on these key NOWS outcomes were significant. These modifiable hospital cultural factors and maternal variables were associated with differences in NOWS outcomes more so than which medication was elected to treat NOWS. This knowledge offers opportunities for improvement in the care of infants affected by NOWS by highlighting modifiable risk factors that can be addressed prenatally and postnatally to improve neonatal outcomes.

Our study is unique in that we compared as-needed methadone and morphine treatment in infants assessed with the ESC Care Tool, whereas previous studies have all used a Finnegan-based or similar (MOTHER scale, a modified Finnegan scoring tool) assessment tools.^21,22  These approaches differ significantly in that ESC is a function-based approach that centers on nonpharmacologic care interventions and the 3 core infant functional domains of eating, sleeping, and ability to console, whereas the Finnegan approach is a numeric scoring system that accounts for a wider range of signs and symptoms of withdrawal with typical cut-off scores for initiating and titrating medication.^16,23,24  Our results differ from the 2 previous RCTs that compared scheduled methadone and morphine dosing, in which NOWS management was guided by a Finnegan-based approach and results revealed that methadone was associated with shorter LOS. We speculate that methadone’s longer half-life compared with that of morphine is beneficial when administered on a standing (scheduled) basis, but differences in half-life may not significantly alter short-term key NOWS outcomes when these pharmacologic agents are administered on a PRN basis.

The total hospital LOS in our study is similar to other studies that used scheduled methadone and morphine, likely related to the number of infants who were converted to scheduled dosing and the high percentage of infants exposed to long-acting MOUD and polypharmacy in our cohort.¹⁶  Those infants who were successfully treated with the PRN approach had a much shorter LOS compared with those converted to standing dosing. As hospitals optimize their PRN treatment approaches, the number of infants who are successfully treated with PRN dosing alone will likely increase.

Interestingly, our results reveal that female sex is associated with decreased LOS, in keeping with current literature that suggests that male infants tend to exhibit more severe NOWS symptomatology and may be more likely to require pharmacologic treatment.²⁵  To our surprise, we also found that female sex was associated with increased odds of conversion to standing dosing and warrants further exploration. Our results are also consistent with other previous studies revealing the association of maternal methadone exposure with longer hospitalizations and length of pharmacologic treatment when compared with buprenorphine and the association of breastfeeding and rooming-in on improved NOWS outcomes.^4,18,26–28  In addition, our study results are consistent with previous reports that have revealed that polypharmacy exposure is associated with worse NOWS hospital outcomes, likely related in part to the medications and in part related to barriers to rooming-in and optimal nonpharmacologic care in these higher-risk dyads.^28,29 

The strengths of this study include the analysis of key outcomes across multiple sites, each having transitioned to the use of PRN pharmacologic agent administration around the same time period. Additionally, our study adjusted for hospital-level random effects that may otherwise affect the outcome. Our results corroborate existing literature regarding the benefits of nonpharmacologic care for infants affected by NOWS and also contribute to the relatively sparse body of literature pertaining to the choice of pharmacologic agent for the treatment of NOWS when administering medication on a PRN basis.

There are important limitations of this study. First, this was not a randomized control trial but rather a retrospective study limited by small sample size and center variability in protocol. Only 1 of the 4 included hospital sites prescribed methadone as the first-line PRN pharmacologic agent whereas the other 3 administered PRN morphine as their first-line agent, potentially biasing study results if there were institution-specific factors that also influenced outcomes. There may exist fundamental differences between the 4 hospital sites in our study because a combination of both academic and community hospitals was included. The criteria for transitioning from PRN to scheduled dosing differed between sites, although in general, they were in the 3 to 5 PRN dose range. Unmeasured confounding variables, including differences between hospitals in the ability of parents to remain at the bedside and the extent of implementation of nonpharmacologic interventions, could have contributed to differences in the need for PRN and the overall pharmacologic treatment rates. Despite this variability, the 4 hospitals included in this study have overall optimized the nonpharmacologic care of infants affected by NOWS through a statewide quality improvement initiative. Key NOWS hospitalization outcomes may differ in other hospital settings in which nonpharmacologic care practices are not as well-established or routinely used. The results of this study may therefore not be generalizable to a larger opioid-exposed neonatal population throughout the United States.

Our study results suggest that nonpharmacologic care practices are of utmost importance in the management of neonates affected by NOWS. However, we are still not able to identify which nonpharmacologic care practices were most efficacious or how they were applied to each eligible neonate both within and across the included hospital sites. Lastly, our study evaluates methadone versus morphine administered on a PRN basis but does not include other pharmacologic agents used for the treatment of NOWS, such as buprenorphine, and outcomes with these alternative agents may differ.

In summary, the study suggests that the choice of morphine versus methadone matters less in impacting NOWS hospitalization outcomes when administered as PRN dosing. Some hospitals may use more PRN doses compared with others, highlighting how hospital cultural factors play a major role in NOWS hospital outcomes. Our results also support existing literature revealing that efforts to optimize nonpharmacologic care practices like breastfeeding and infant care location outside of the NICU may help improve NOWS outcomes more than the administration of a specific pharmacologic agent, particularly for hospital sites that have a well-established and routine implementation of nonpharmacologic NOWS care bundles. The study results also reveal that there is a significant association between modifiable maternal factors such as prenatal nonprescribed drug exposure and benzodiazepine exposure and worse NOWS outcomes. This knowledge provides an opportunity to identify which infants and families could benefit from additional prenatal efforts for engagement, and postnatal efforts to improve first-line nonpharmacologic treatment to improve outcomes.

In conclusion, our study provides additional insight into NOWS pharmacologic agent choice provided on a PRN basis. The choice of pharmacologic agent (methadone versus morphine) may be less clinically relevant compared with an awareness of antenatal factors like nonprescribed drug exposure, benzodiazepine or SSRI exposure, and the utilization and optimization of nonpharmacologic care practices to improve hospitalization outcomes of neonates affected by NOWS. Future studies should continue to optimize and standardize PRN dosing protocols to increase the number of neonates who can be successfully treated with this approach.