Clonidine as Monotherapy for Neonatal Opioid Withdrawal Syndrome: A Randomized Trial

This study (https://clinicaltrials.gov/ct2/show/NCT03396588) had approval from the institutional review board (IRB) and a certificate of confidentiality from the National Institute on Drug Abuse. Parent(s) gave informed consent for study participation. This study was an intention-to-treat, double-blind trial, with treatment randomized between clonidine and morphine, carried out at the Kentucky Children’s Hospital. Primary outcomes included LOT and post-treatment neurobehavioral status, which is a reported predictor of childhood outcomes.⁵¹ 

Research nurses screened all admissions for prenatal substance exposure at the well-baby nursery and the NICU. Enrollment criteria included (1) gestational age ≥35 weeks, (2) any prenatal opioid exposure with or without other substances from the mother reporting opioid use or positive toxicology results, (3) ≤7 days postnatal age, (4) not unlikely to survive, (5) no medical condition other than NOWS, (6) no seizures, (7) no prenatal cocaine exposure, (8) meets treatment criteria for NOWS, and (9) informed consent. Prenatal cocaine exposure was excluded because tone abnormalities have been attributed to cocaine central nervous system effects rather than drug withdrawal.^52,53 

All infants had non-pharmacological intervention (swaddling, low noise, and lighting environment, rooming in, infant massage, etc).^54–56  Throughout the hospital stay, infants were monitored using the Finnegan Neonatal Abstinence Scoring System.⁵⁵  The Finnegan Neonatal Abstinence Scoring System scores (FS) guided the initiation and monitoring of treatment, allowing for comparison with other treatment studies.^45,57,58  Clonidine or morphine was started when an infant had 3 consecutive FSs ≥8 or 2 consecutive FSs ≥12 with the evaluation and approval of the attending physician. Treatment assignment was obtained through the Investigational Drug Service of the Center for Clinical and Translational Science. Clinical pharmacists prepared and dispensed the study drug, with clonidine and morphine having identical appearances. Block randomization allowed for balanced enrollment in each treatment arm. Families, providers, and research staff were masked to the treatment assignment. Our initial drug dosage (clonidine, 1 µg/kg/dose or morphine, 60 µg/kg/dose) approximated our previous study’s receiver operating characteristics that provided the best control of signs⁵⁰  and was given every 3 hours concomitant with the infant’s feeding schedule. The initial clonidine dose was within the range given in other studies^42–46  and supported by published population pharmacokinetics.⁵⁹ 

The dose for either drug was increased by 25% of the initial dose every 12 to 24 hours until FSs were consistently <8, to a maximum of 2 μg/kg/dose for clonidine and 120 µg/kg/dose for morphine. The maximum clonidine dose of 16 μg/kg/day was slightly higher than reported by Agthe et al⁴⁵  but within the dose range reported by O’Mara et al.⁶⁰ 

After 24 to 48 hours of stabilization, dose weaning began by 10% of the maximum dose every 24 hours. When FSs increased during weaning, the dose was increased to the previous dose; weaning resumed after 24 to 48 hours. If no improvement was noted at the maximum dose, phenobarbital was added at a loading dose of 10 mg/kg and maintenance dose of 2.5 mg/kg q 12 hours.⁶¹  Some attending physicians preferred clonidine or morphine as an adjunct. Thus, infants were allowed to receive either morphine or clonidine as the secondary drug, which the infant was not randomly assigned to receive. Masking was maintained. After stabilization after the addition of the adjunct drug, the primary (first) drug was weaned daily by 10% of the maximum dose. After discontinuation of the primary drug, adjunct drug weaning followed. After 48 hours of observation without medication, infants were discharged. Infants on phenobarbital were discharged with a weaning schedule over 3 weeks.

Drug use during pregnancy was obtained through maternal interviews by research staff, who also abstracted medical records for demographic information, including race and ethnicity, maternal pregnancy history, drug use, MOUD, delivery information, FSs, dose changes, duration of treatment, and drug effects, conditions, complications for reporting to the IRB and Drug Safety Monitoring Board (DSMB) of the Center for Clinical and Translational Science. Data entry and management were conducted by using Research Electronic Data Capture tools (REDCap 11.4.2 - 2021 Vanderbilt University).^62,63 

The NICU Network Neurobehavioral Scale (NNNS)^64,65  requires eliciting responses from a sleep or quiet state to crying (behavioral state 6). Infants with severe withdrawal are typically unable to complete the assessment. Therefore, after withdrawal signs improved and dose weaning began, trained and certified research nurses, who were blinded to treatment assignment, performed the initial assessment; this was repeated at treatment completion. The NNNS assesses neurologic, behavioral, and stress/abstinence components from the Brazelton Scale⁶⁶  and Finnegan’s work,⁶⁷  yielding 13 summary scores:⁶⁵  habituation, attention, handling, quality of movement, regulation, non-optimal reflexes, asymmetry total/reflexes, stress/abstinence, arousal, hypertonicity, hypotonicity, excitability, and lethargy.

When designing this study, we anticipated recruiting ∼200 infants. Assuming 20% attrition (80 infants in each treatment group), there would be at least 80% power to detect a difference with an effect size of at least 0.45 when using a 2-sided test with a 5% significance level. This effect size was exceeded for some outcomes in the pilot data.⁵⁰  An interim analysis was conducted in 2020 for the DSMB. At that time, 94 subjects had been randomly assigned. The study was allowed to continue.

Based on Consolidated Standards of Reporting Trials (CONSORT) standards, baseline demographic variables were summarized for morphine and clonidine treatment.^68,69  Frequencies (%) are presented for categorical variables and either mean (SD) or median (min, max) for continuous variables. For outcome results, we present 95% confidence intervals (CIs).

A log-rank test was used to compare treatment arms regarding LOT and LOS.⁷⁰  To use data from infants withdrawn from the study, the time of withdrawal was specified as the censoring time. Cox proportional hazards regression modeling was used in a follow-up analysis using treatment and selected risk factors associated with NOWS severity (birth weight, gestational age, gender, cesarean delivery, maternal psychiatric disorder, inborn status, MOUD, other opioid, and non-opioid substance use).^{32,51,71–74}  Because smoking was highly prevalent and alcohol use was infrequent, they were not included in the model. In post hoc analyses, a negative binomial regression was used to model LOT for 115 infants, excluding those withdrawn from the study. Empirical SE estimates were employed to ensure valid inference.

Treatment arms were compared regarding the need for adjunct medication using frequencies (%) and χ² tests. Of 5 infants withdrawn, only 1 received adjunct medication; the other 4 were observed for <2 days. Utilizing data from all 120 infants, we assumed that the 4 infants did not require adjunct treatment. We also removed these 4 infants in a follow-up analysis. The remaining 116 infants were used to fit a logistic regression model, from which adjusted odds ratios were derived.

Wilcoxon rank tests were used to compare treatment groups regarding the timing between initial and final NNNS assessments and the NNNS scores (initial, final, and change). P values were adjusted using the false discovery rate method because of multiple comparisons across 13 summary scores.⁷⁵ 

Intention-to-treat analyses used all available data based on treatment assignment. All tests were 2-sided, with a statistical significance of P <.05. Analyses were conducted in SAS v9.4 (SAS Institute Inc 2013. SAS/ACCESS® 9.4 Interface to ADABAS: Reference. Cary, NC: SAS Institute Inc.).

The enrollment period started in December 2017 and ended in February 2022, at funding completion. In 2020 and 2021, coronavirus disease 2019 restrictions over face-to-face encounters with subjects resulted in decreased enrollment. Enrollment continued through phone consent and follow-up post-discharge through virtual visits.

The CONSORT diagram is in Fig 1. Infants with prenatal substance exposure (n = 1107) were assessed for eligibility. There were 155 whose parents gave consent, but 35 did not meet the treatment criteria. The remaining 120 infants were randomly assigned to receive clonidine (n = 60) or morphine (n = 60). Five infants were withdrawn from the study (clonidine [n = 2] and morphine [n = 3]). Four of 5 infants were withdrawn after 1 to 2 days of treatment; 1 randomly assigned to receive clonidine was withdrawn after 10 days of treatment because of abnormal chromosome results.

Table 1 reveals characteristics of mothers and infants by treatment with no post hoc comparison, per the CONSORT statement.^68,69  A total of 52 (87%) and 49 (82%) mothers in the clonidine and morphine groups, respectively, were receiving MOUD. Despite this, the use of other substances was prevalent (Table 2) based on maternal self-report and toxicology results.

Before treatment, the median (95% CI) of the maximum FS was similar in clonidine- and morphine-treated infants (14.5 [14.1–15.6] and 14 [13.9–15.5], respectively). FS decreased subsequently. By the third day of treatment, the maximum FS was significantly lower (P < .001) in the morphine-treated group, with a median (95% CI) of 8.0 (7.3–8.6) versus 11.5 (11.0–12.5) in the clonidine group. In Fig 2A–B, there are similar patterns of mean FS (95% CI) over time for the treatment groups and subgroups with monotherapy. Mean FS scores were similar at treatment initiation, with a more pronounced drop by the second day of treatment with morphine treatment. After 6 days, the mean daily scores of the clonidine group trended lower than the morphine group, although the CIs overlapped.

The LOT did not differ significantly between treatments (P = .48); the medians (95% CI) were 17 (15–19) and 15 (13–17) days, respectively, for the clonidine and morphine groups. Neither was there a significant difference in LOS between clonidine and morphine groups, with medians (95% CI) of 22 (20–24) and 19 (17–21) days, respectively (P = .30). In those receiving a single therapy, the LOT was similar for clonidine and morphine subgroups, with medians (95% CI) of 14 (12–16) days and 14 (12–17) days, respectively (P = .59). The mean (SD) total morphine dose per patient in the clonidine arm was 5.5 (13.2) mg and 9.1 (19.8) mg in the morphine arm. The clonidine-treated infants who also received morphine had a mean (SD) of 27.7 (16.0) mg total morphine dose per patient; the morphine treatment duration was longer, starting a few days from clonidine initiation and weaning beginning only after clonidine discontinuation.

In an unadjusted model, the hazard ratio (HR) for clonidine to morphine was estimated to be 0.88 (95% CI: [0.61–1.28], P = .50). In the adjusted model (Table 3), the results were similar, (HR = 0.82, 95% CI: [0.55–1.23], P = .33). A heavier birth weight was associated a with longer LOT (HR = 0.95, 95% CI: [0.91–0.99], P = .01). Exposure to any other opioids predicted a shorter LOT, but not significantly when exposures were only based on toxicology results.

When excluding infants who were withdrawn from the study, the results were similar (Supplemental Table 5). The unadjusted negative binomial model showed an estimated mean duration of treatment of clonidine to be 18.8 days and 18.0 days for morphine, corresponding to a ratio of 1.05 (95% CI: 0.87–1.26).

More clonidine-treated infants required adjunct therapy than the morphine-treated infants, as follows: 27/60 (45%) and 6/60 (10%), respectively (P <.001). The results were similar when excluding the withdrawn subjects (P <.001). The adjusted odds for adjunct therapy were higher for clonidine (odds ratio = 8.85, 95% CI: [2.87–27.31], P <.001; Table 3). No other factor except heavier birth weight was associated with adjunct therapy (odds ratio = 1.17, 95% CI [1.04–1.31]).

For the clonidine group (n = 56), the median (min, max) interval between treatment initiation and the initial NNNS examination was 5 (1–16) days (1 outlier), which is significantly longer compared with 3 (1–9) days in the morphine-treated (n = 57) group (P <.001). For a post hoc comparison of subgroups, clonidine only (n = 31) and morphine only (n = 51), the median (min, max) intervals, respectively, were 3 (1–8) days and 3 (1–9) days (P = .52).

The interval between treatment initiation and the final NNNS examination did not differ significantly between treatment groups; for clonidine (n = 54), the median (min, max) was 16.5 (8–48) days, and for morphine (n = 51), it was 15 (7–50) days (P = .24). For monotherapy subgroups, clonidine (n = 32) and morphine (n = 45), the intervals were 14 (8–24) and 14 (7–32), respectively, and P = .94.

The initial, final, and change in the NNNS summary scores of the treatment groups are shown in Table 4. The results are similar when comparing monotherapy subgroups (Supplemental Table 6). At the initial testing, clonidine-treated babies scored worse on arousal, hypertonicity, and stress abstinence. However, between the initial and final assessments, with the duration being similar between groups, the clonidine-treated infants showed significant improvement (less handing, excitability, stress abstinence, and improved arousal and regulation); thus, after treatment completion, clonidine-treated infants did as well as those who were morphine-treated regarding neurobehavioral performance.

Conditions monitored and reported to both the DSMB and IRB are in Supplemental Table 7. These conditions were rare and may not be related to treatment.

In this randomized trial comparing the clonidine and morphine treatment of NOWS, we found no significant differences in LOT and LOS. However, nearly one-half of the infants treated with clonidine received adjunct therapy. The FS decreased with the initiation of treatment, but the decrease in scores in clonidine-treated babies lagged behind that of morphine-treated infants by 1 to 2 days, resulting in the initiation of adjunct therapy. Initially, clonidine-treated infants had worse neurobehavioral performance scores but were no longer different from morphine-treated infants at treatment completion within similar intervals between the initial and final assessments.

The lack of difference in LOT between groups is notable, considering a larger proportion of clonidine-treated babies receiving adjunct therapy and adjustment for multiple factors. We were able to successfully provide an opioid-sparing medication-assisted treatment regimen to a cohort of infants with significant NOWS. This use of non-opioids may be extended for conditions other than NOWS, such as for the relief of chronic pain or in postoperative states.

A high proportion of clonidine-treated infants needing adjunct therapy is a concern. However, this could be addressed by adjusting the dosing schedule. On the basis of our pilot trial results of conservatively starting clonidine dosing at 0.5 µg/kg,⁵⁰  published data,⁷⁶  and a plan to monitor possible adverse events, we started clonidine at 1 µg/kg. Considering that the estimated half-life of clonidine in a neonate is 14 hours,⁵⁹  compared with 6.5 hours^77,78  for morphine, it is reasonable to assume that the maximum pharmacologic effects would be near the time of achieving steady-state, which is typically 5-fold the agent’s half-life, or ∼70 hours for clonidine and ∼32 hours for morphine. This is consistent with the lag in the decrease of FS by 1 to 2 days with clonidine treatment. Conversely, the subsequent trend of lower mean FS with clonidine treatment probably reflects achieving a steady state. This delayed onset of efficacy for clonidine likely necessitated dose escalations and adjunct medication. We are currently examining our pharmacokinetics data to optimize the dosing regimen with clonidine to shorten the time to steady state.

The worse neurobehavioral performance of the clonidine-treated infants compared to the morphine group on the initial assessment may be explained by the delayed or lesser effectiveness of the clonidine response or not having a designated specific time point for the initial assessment. The improvement in neurobehavioral performance with clonidine treatment is reassuring; neurobehavioral functioning is reported to predict early childhood outcomes in drug-exposed infants.⁵¹  Long-term follow-up is needed to understand the implications of NOWS and its various treatment modalities.

This study was not powered to show equivalence or non-inferiority. Having this limitation, the lack of significant difference may reflect a type II hypothesis testing error, which would become less probable with a larger sample size. The coronavirus disease 2019 pandemic impacted our plans to enroll a larger sample and add another study site. Nonetheless, the number of infants randomly assigned in our study was similar to that of other multicenter treatment trials.^58,79  Being a single study site limits the generalizability of findings. However, we provide information on infants from rural areas. Some mothers did not room in with their babies, and many did not initiate breastfeeding; these affect non-pharmacological care and LOS.^80,81  Although a high proportion of mothers were on MOUD, the use of other substances was common. Polypharmacy may also influence NOWS treatment.⁸²  Ascertaining the effects of polydrug use will require a larger number of participants.

In summary, in a randomized trial, the length of pharmacologic treatment and the final assessment of neurobehavioral performance were not significantly different between the clonidine- and morphine-treated groups. Future studies are needed to investigate the optimal dose and frequency of clonidine administration for improved efficacy and the decreased need for adjunctive therapy in NOWS.

We thank the following for their contribution to the completion of this research: Gail L. Camp, Heather Collins, Susan DeGraff, Caitlin Dunworth, Karen Garlitz, Pui Man (Julia) Ho, Carrie Hobbs, Jamie Horn, Seth Larkin, Alisa McKinney-Whitlock, Leanne Pollard, Brandon Schanbacher, Ashley Stratton, and Amanda Wilburn. We also thank anonymous peer reviewers for their comments, which informed the revision of this work.

CI

confidence interval

CONSORT

Consolidated Standards of Reporting Trials

DSMB

Drug Safety Monitoring Board

FS

Finnegan Neonatal Abstinence Scoring System score

HR

hazard ratio

IRB

institutional review board

LOS

length of stay

LOT

length of treatment

MOUD

medication for opioid use disorder

NAS

neonatal abstinence syndrome

NOWS

neonatal opioid withdrawal syndrome

NNNS

NICU network neurobehavioral scale