Non‐pharmacological care for opioid withdrawal in newborns

Summary of findings 1. Non‐pharmacological treatment compared to placebo for infants at risk for, or having early symptoms consistent with, opioid withdrawal

Outcomes	*Anticipated absolute effects (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)
Non‐pharmacological treatment compared to placebo for infants at risk for, or having early symptoms consistent with, opioid withdrawal
Patient or population: infants at risk for, or having early symptoms consistent with, opioid withdrawal Setting: newborn or neonatal unit, USA and UK Intervention: non‐pharmacological treatment Comparison: placebo
Outcomes	Risk with placebo	Risk with non‐pharmacological treatment	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)
*Primary outcome measures*
Length of hospitalization (days)
– Modifying environmental stimulation	Mean was 11.5 days (SD 3.4)	MD 1 lower (2.82 lower to 0.82 higher)	—	30 (1 RCT)	⊕⊝⊝⊝ Very low^a
– Feeding practices	—	—	—	—	—
– Support of the mother‐infant dyad	Mean was 19.4 days (SD 13)	MD 8.9 lower (19.84 lower to 2.04 higher)	—	14 (1 RCT)	⊕⊝⊝⊝ Very low^b
Pharmacological treatment with ≥ 1doses of opioid or sedative medication
– Modifying environmental stimulation	Study population		RR 1.00 (0.86 to 1.16)	92 (3 RCTs)	⊕⊕⊝⊝ Low^c
– Modifying environmental stimulation	844 per 1000 (533 to 1000)	851 per 1000 (533 to 1000)	RR 1.00 (0.86 to 1.16)	92 (3 RCTs)	⊕⊕⊝⊝ Low^c
– Feeding practices	Study population		RR 0.92 (0.63 to 1.33)	49 (1 RCT)	⊕⊝⊝⊝ Very low^d
– Feeding practices	727 per 1000	667 per 1000	RR 0.92 (0.63 to 1.33)	49 (1 RCT)	⊕⊝⊝⊝ Very low^d
– Support of the mother‐infant dyad	Study population		RR 0.50 (0.13 to 1.90)	14 (1 RCT)	⊕⊝⊝⊝ Very low^b
– Support of the mother‐infant dyad	571 per 1000	286 per 1000	RR 0.50 (0.13 to 1.90)	14 (1 RCT)	⊕⊝⊝⊝ Very low^b
Secondary outcome measures**
NICU admission
– Support of the mother‐infant dyad	Study population		RR 0.50 (0.13 to 1.90)	14 (1 RCT)	⊕⊝⊝⊝ Very low^b
– Support of the mother‐infant dyad	571 per 1000	286 per 1000	RR 0.50 (0.13 to 1.90)	14 (1 RCT)	⊕⊝⊝⊝ Very low^b
Days to regain birthweight
– Feeding practices	Mean was 14.7 days (SD 2.84)	MD 1.1 lower (2.76 lower to 0.56 higher)	—	46 (1 RCT)	⊕⊝⊝⊝ Very low^e
Weight nadir (% weight loss)
– Modifying environmental stimulation	Mean was 5.5% weight loss	MD 0.28 lower (1.15 lower to 0.59 higher)	—	194 (1 RCT)	⊕⊝⊝⊝ Very low^f
– Feeding practices	Mean was 9.4% weight loss	MD 0.8 lower (2.24 lower to 0.64 higher)	—	46 (1 RCT)	⊕⊝⊝⊝ Very low^d
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). Only interventions that reported on secondary outcomes are included in the 'Summary of findings' table. CI: confidence interval; MD: mean difference; NICU: neonatal intensive care unit; RCT: randomized controlled trial; SD:** standard deviation.
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aDowngraded three levels for very serious concerns about risk of bias (due to unblinded study, concern for performance, detection, and attrition bias) and serious concern for imprecision (due to wide CIs). ^bDowngraded three levels for very serious concerns about risk of bias (due to unblinded study, concern for performance, detection, and attrition bias) and very serious concern for imprecision (due to very wide CIs). ^cDowngraded two levels for very serious concerns about risk of bias (due to unblinded study, concern for performance, detection, and attrition bias). ^dDowngraded three levels for very serious concerns about risk of bias (due to concern for attrition and reporting bias), serious concern about imprecision (due to wide CIs), and plausible confounding that would reduce demonstrated effect. ^eDowngraded three levels for very serious concerns about risk of bias (due to concern for attrition and reporting bias), very serious concern about imprecision (due to very wide CIs), and plausible confounding that would reduce demonstrated effect. ^fDowngraded three levels for very serious concerns about risk of bias (due to unblinded study, concern for performance and detection bias) and serious concern for imprecision (due to wide CIs).

Background

Substance use, both prescribed and non‐prescribed, is an international public health concern. Prevalence is increasing in many areas of the world. Substance use by women of childbearing age contributes to increasing neonatal abstinence syndrome (NAS). Although mothers may have multiple substance exposures, opioid exposure constitutes the majority of cases described historically as NAS (Hudak 2012). Neonatal opioid withdrawal syndrome (NOWS) is a newer term describing the subset of NAS related to opioid exposure (Kelly 2020). For the purposes of this review, we will use the broader term of NAS to include the broader definition used in much of the literature.

Incidence of NAS varies by location, and rising rates are not universal. The incidence of NAS in Australia and England is 2.7 per 1000 hospital births and stable (Davies 2016). The incidence of NAS in the USA and Canada is rising. Incidence increased four‐fold in the USA between 2003 and 2012 to 5.8 per 1000 hospital births, then rose further to 7.0 per 1000 hospital births in 2016 (Corr 2017; HCUP 2019; Patrick 2015a). Incidence of NAS in Canada rose 21% between 2013 and 2017 to 5.2 per 1000 hospital births (CIHI 2018). There is also regional variability within countries. In New South Wales, Australia, incidence of NAS was 3.18 per 1000 hospital births in 2011 (Uebel 2016). Within the USA, there is a wide range in incidence of NAS, ranging from 0.7 to 33.4 per 1000 hospital births (Ko 2016). In 23 USA hospitals, NAS affected a range from less than 1% to 18% of neonatal encounters (Milliren 2018). Genetic, socioeconomic, racial, and local factors including legislation, culture, and medical practices contribute to the variability (Li 2009). Substance use affects all socioeconomic and racial groups, but is disproportionately reported in women of color and low socioeconomic status (Paltrow 2013). Incidence of NAS is 14.4 per 1000 hospital births in the subgroup of low‐income USA infants with public insurance (Winkelman 2018). Punitive state legislation policies are also associated with higher incidence of NAS (Faherty 2019). Under‐recognition and variable use of diagnostic codes may also affect the measured incidence of NAS (Burns 2007; Goyal 2020).

Care for infants with NAS in the neonatal period, and later in childhood, place demands on the medical system, social welfare system, educational system, and society at large. In 2013, infants with NAS occupied 4% of neonatal intensive care unit (NICU) beds in the USA, an increase from 0.6% in only nine years (Tolia 2015). Over 95% of these patients are withdrawing from prenatal exposure (Patrick 2012). Infants with NAS are at increased risk of morbidity or mortality in the neonatal period (Lisonkova 2019; Patrick 2012). Parental substance use is associated with increased child welfare involvement (O'Donnell 2009; Patrick 2019). Later in childhood, children with a history of NAS have an increased risk of rehospitalization (Uebel 2015). Non‐randomized studies demonstrate poorer cognitive and behavioral outcomes through childhood and into high school (Lee 2020; Yeoh 2019). Importantly, many of these associations are based on non‐randomized studies with both recognized and unrecognized complex confounding factors (Larson 2019). Further study is needed to distinguish causation related to opioid exposure versus other social, environmental, or biological factors.

Description of the condition

The contemporary definition of NAS is broad, and includes symptoms associated with prenatal exposure to many substances, both prescribed and non‐prescribed. Symptoms related to withdrawal from opioids, such as methadone, buprenorphine, heroin, oxycodone, and other prescribed and non‐prescribed opioids, may also be described as NOWS. Opioid withdrawal is a unique subset of NAS characterized by common symptomatology and is more likely to respond to postnatal treatment with opioids (Hudak 2012). Polysubstance use is common among opioid users and is associated with more severe NAS (Wachman 2018a). Non‐opioid substance exposures may be associated with NAS due to ongoing toxic effects of exposure (e.g. cocaine) or due to withdrawal following birth (e.g. benzodiazepines). Substances associated with NAS include benzodiazepines, selective serotonin reuptake inhibitors (SSRIs), and tobacco, among other substances. While exposure may occur in isolation or in combination with other substances, most documented NAS is at least partly due to opioids (Hudak 2012).

Infants express varied signs of withdrawal from opioids and other substances following birth. Symptoms are those of dysregulated autonomic, gastrointestinal, and central nervous systems. The most severe manifestations of withdrawal may include seizures, which are likely rare, and severe weight loss. Symptoms of opioid withdrawal peak in the first week of life due to cessation of opioid transfer through the placenta. Infants need medical monitoring and treatment during this period. Symptoms are more severe with maternal methadone maintenance, smoking, and polysubstance exposure (Brogly 2014; Patrick 2015b; Wachman 2018a). Symptoms are less severe in preterm infants; this may be related to a different physiologic response to opioids or a different manifestation of withdrawal in the preterm population (Lemon 2018).

The American Academy of Pediatrics recommends standardized assessment of symptom severity (Hudak 2012). Published standardized withdrawal assessment tools include the Finnegan Neonatal Abstinence Scoring Tool (score range 0 to 62) (Finnegan 1975), Lipsitz Neonatal Drug Withdrawal Scoring System (score range 0 to 20) (Lipsitz 1975), Neonatal Narcotic Withdrawal Index (score range 0 to 14) (Green 1981), Neonatal Withdrawal Inventory (score range 0 to 19) (Zahorodny 1998), and Maternal Opioid Treatment: Human Experimental Research (MOTHER) NAS scale (score range 0 to 42) (Jones 2010). Several recent efforts have been made to simplify scoring. Examples include Finnegan Neonatal Abstinence Syndrome Scale – Short Form (score range 0 to 16) (Maguire 2013), MOTHER NAS Short Form (score range not reported) (Jones 2016a), simplified Finnegan Neonatal Abstinence Scoring System (score range 0 to 23) (Gomez Pomar 2017), a new scoring system for NAS (score range 0 to 27) (Kocherlakota 2020), and the eight‐item simplified Finnegan Neonatal Abstinence Scoring System (score range 0 to 8) (Devlin 2020). Each tool assesses a combination of symptoms of opioid or non‐opioid withdrawal; however, most tools were designed in a time when single drug use was common, primarily heroin. Most hospitals use a version of the Finnegan Neonatal Abstinence Scoring Tool (Mehta 2013; Sarkar 2006; Snowden 2019); however, the pharmacological treatment threshold is not validated and the tool has poor psychometric properties and poor internal consistency (Jones 2016b). Newer approaches in the USA shift from reliance on a threshold score to initiate pharmacological treatment to an approach based on a care pathway. One example focuses on early modulation of symptoms with a standard non‐pharmacological care bundle including family involvement and a more nuanced evaluation of critical functional symptoms including the infant's ability to eat, sleep, and be consoled (Grossman 2018; Wachman 2018b).

The first‐line treatment for NAS is non‐pharmacological care (Hudak 2012; Mangat 2019). Severe symptoms despite maximal non‐pharmacological treatment mark a threshold for pharmacological treatment. Though norms exist, scores are often subjective, and treatment thresholds established based on historical norms are not validated. Pharmacological treatments that reduce withdrawal symptoms include opioid or sedative medications, or both. Morphine is the most common opioid treatment; methadone and buprenorphine are also used (Patrick 2016; Snowden 2019). Adjunctive therapies may include phenobarbital and clonidine to address polysubstance exposure. Long‐term developmental impact of pharmacological treatment is unclear. Immediate adverse effects of pharmacological treatment include respiratory depression and sedation. Most hospitals care for infants receiving sedating medications on a cardiac monitor in the intensive care unit (ICU) (Milliren 2018). Unfortunately, ICU care separates infants from caregivers and often is a high stimulus environment. This interferes with bonding, caregiver self‐efficacy, and full non‐pharmacological treatment, potentially affecting severity of NAS symptoms.

The approach to care for infants with NAS has wide variation (Bogen 2017; Mehta 2013; O'Grady 2009; Patrick 2016; Sarkar 2006; Snowden 2019). Surveys show differences in prenatal counseling, written policies, staff training, threshold for pharmacological treatment, weaning protocols, breastfeeding support, and location of care, among other aspects of care (Bogen 2017; Mehta 2013; Patrick 2016; Snowden 2019). Most hospitals use a standardized withdrawal assessment tool; however, hospitals have differing treatment protocols. Almost half lack a written policy for management of opioid‐exposed newborns (Patrick 2016; Sarkar 2006). There is no clear standard for non‐pharmacological care or pharmacological treatment thresholds (Mehta 2013). Non‐pharmacological treatment and pharmacological treatment thresholds are not well studied and likely vary between centers. Most importantly, efforts to standardize hospital policies are associated with improved outcomes (Patrick 2016).

Outcomes such as pharmacological treatment rate and length of hospital stay also vary widely. The percent of infants who receive pharmacological therapy ranges from 13% to 90% (Milliren 2018). The mean length of stay at some centers in the USA is as low as 5.9 days, while the national mean is 16.9 days (Grossman 2017; Patrick 2015a).

Variations in diagnosis and management of NAS represent an important opportunity to improve care. Multicenter quality improvement collaboratives in the USA show that increased standardization of care can reduce length of hospital stay (Hall 2015; Patrick 2016). Emerging evidence suggests that standardization of non‐pharmacological care also reduces length of hospital stay and use of pharmacological treatment (Grossman 2017; Holmes 2016; Wachman 2018b).

Description of the intervention

Non‐pharmacological treatments focus on minimizing dysregulation and maximizing infant functioning. Treatments are heterogeneous interventions within three main categories of care:

modifying environmental stimulation;
feeding practices; and
support of the mother‐infant dyad.

Interventions may occur alone, or in combination with other non‐pharmacological and pharmacological treatments. The 'dose' of non‐pharmacological intervention is difficult to assess, and may be related to timing, intensity, quality, consistency, and adequacy of interventions among other features.

Modifying environmental stimulation

Treatments addressing environmental stimulation reduce negative stimuli and promote positive experiences. Gentle handling and maintaining a low‐light, quiet environment reduce negative stimulation. Infant containment or swaddling, bedding choice including vibrating beds, and infant positioning may also soothe infants. Positive experiences may include non‐nutritive sucking and bathing. Aromatherapy, music therapy, massage, and acupuncture/acupressure are therapies that may also calm infants through stimulation.

Feeding practices

Disorganized feeding and weight loss are common in infants withdrawing from opioids. In general, breast milk and breastfeeding reduce withdrawal symptoms (Jansson 2009). Small, frequent feedings and infant‐led feeding are often helpful in improving the transfer of food. Higher‐calorie feedings or tube feedings may reduce weight loss and improve satiety in these infants. Non‐maternal or maternal breast milk or low‐lactose formula may reduce gas and feeding discomfort.

Support of the mother‐infant dyad

Treatments also address the health and support of the mother‐infant dyad as a unit (Krans 2019; Velez 2009). Parental presence alone is an important aspect of non‐pharmacological care (Howard 2017). Rooming in (where the baby's crib is kept by the side of the mother's bed) and skin‐to‐skin care promote mother‐infant bonding and infant regulation. Direct parental support promotes parental well‐being. Well caregivers are better able to recognize and respond to infant cues. Parental supports enhance comfort and plan for respite. Examples include 'cuddler' programs (where cuddling the baby is maximized) and designated spaces in the hospital to promote parent comfort. Staff contribute to parental well‐being by using trauma‐informed care principles. A comprehensive biopsychosocial needs assessment may also improve parental well‐being and prompt referral for comprehensive integrated care (Krans 2019). Needs may include addiction, social, mental health, legal, and custody supports.

How the intervention might work

Many non‐pharmacological practices improve autonomic regulation for infants and could mitigate the autonomic dysregulation noted in infants with NAS.

Environmental changes and feeding practices, individually and in combination, can affect physiologic parameters in infants. Facilitated tucking by holding the infant in a gently flexed position lowers stress‐related behaviors and vital sign changes in infants (Liaw 2012). Human touch through massage improves weight gain, growth, and sleep in infants (Juneau 2015). Non‐nutritive sucking and acupressure are associated with reduced infant pain response (Chen 2017; Pinelli 2002). Breastfeeding and breast milk odor are each associated with reduced infant pain response (Carbajal 2003; Gray 2002; Zhang 2018). Breastfeeding, especially combined with maternal holding, is associated with widespread cortical activity distinct from that of glucose exposure alone (Bembich 2018).

Maternal interaction and presence have many unique benefits for the infant, likely related to complex and multifactorial sensory stimuli involved in interactions within the mother‐infant dyad. Skin‐to‐skin contact between caregiver and infant (also termed 'kangaroo care') is associated with a reduced infant pain response (Gray 2000; Johnston 2017; Pandita 2018; Shah 2012). Mother's presence and tactile contact improve visual attention in term infants (Arditi 2006). Maternal well‐being and functioning appear to affect infant regulation. Sensitive maternal caregiving as measured by the Emotional Availability Scales is associated with lower infant cortisol and improved infant regulation (Kaplan 2008). Relationships between parents, specifically the levels of avoidance and dyadic adjustment, are associated with infant vagal tone (Graham 2010).

This literature suggests that non‐pharmacological measures affect autonomic functioning and are likely to attenuate withdrawal symptoms due to dysregulation in NAS.

Why it is important to do this review

Other Cochrane Reviews address the evidence for, and comparative effectiveness of, pharmacological treatments including opiates, sedatives, and naloxone treatments for opioid‐exposed newborns (Moe‐Byrne 2018; Osborn 2010a; Osborn 2010b). The reviews conclude that initial pharmacological treatment of neonatal opioid withdrawal should be an opiate. Opiates reduce the time to regain birth weight but increase the length of hospital stay. Adding a sedative may further reduce symptoms. Given the methodologic limitations of the included studies, further firm recommendations are not made.

Non‐pharmacological interventions are not assessed or controlled in most studies focused on pharmacological treatments for NAS. The use and 'dose' of non‐pharmacological care may be an important confounder in these studies. Standardized non‐pharmacological care may improve care outcomes (Grossman 2017; Holmes 2016; Wachman 2018b). Evidence for non‐pharmacological interventions will guide future efforts to improve care. This review evaluated the current literature and gaps in research regarding non‐pharmacological care. Acupuncture and acupressure will be addressed in a separate Cochrane Review, and was not included in this review.

Objectives

Methods

Criteria for considering studies for this review

Types of studies

We included randomized controlled trials (RCTs), quasi‐RCTs, and cluster trials in our formal analysis (see Selection of studies). We excluded cross‐over trials.

Types of participants

Inclusion criteria

Comparison 1

Term (37 weeks' gestation or greater) and late preterm infants (34 weeks' gestation to 37 weeks' gestation) with known or suspected prenatal opioid exposure or having opioid withdrawal in the first seven days of life.

Comparison 2

Term (37 weeks' gestation or greater) and late preterm infants (34 weeks' gestation to 37 weeks' gestation) on opioid treatment for opioid withdrawal in the first 28 days of life.

Opioid withdrawal may have been defined by the presence of symptoms consistent with opioid withdrawal or an elevated score on a standardized withdrawal assessment tool.

Symptoms consistent with opioid withdrawal are those of dysregulated autonomic, gastrointestinal, and central nervous systems. Symptoms included, but were not limited to: increased crying, irritability, inability to sleep, tremors, increased tone, myoclonic jerks, fever, sweating, rapid breathing, labored breathing, sneezing, yawning, unco‐ordinated feeding, excessive sucking, vomiting, loose stools, and excessive weight loss.

Scores are often used to quantify the symptoms and provide information to guide pharmacological treatment. Elevated scores on standardized withdrawal assessment tools include Finnegan Neonatal Abstinence Scoring Tool score above 8 (Finnegan 1975), Lipsitz Neonatal Drug Withdrawal Scoring System score above 4 (Lipsitz 1975), Neonatal Narcotic Withdrawal Index score above 5 (Green 1981), Neonatal Withdrawal Inventory score above 8 (Zahorodny 1998), MOTHER NAS score above 9 (Jones 2010), Finnegan Neonatal Abstinence Syndrome Scale – Short Form score above 8 (Maguire 2013), or a score exceeding the stated threshold on any novel validated assessment tool.

Exclusion criteria

Preterm infants less than 34 weeks' gestation.
Critically ill term and late preterm infants (greater than 34 weeks' gestation) with medical comorbidities unrelated to opioid withdrawal. Medical comorbidities included but were not limited to the need for mechanical ventilation (positive pressure ventilation), culture‐confirmed infection, necrotizing enterocolitis, major congenital anomaly, and status postmajor surgical intervention.

Types of interventions

Comparison 1

Trials comparing single or bundled non‐pharmacological interventions to no non‐pharmacological treatment or different single or bundled non‐pharmacological interventions in infants at risk for, or having early symptoms consistent with, opioid withdrawal as defined above.

Comparison 2

Trials comparing single or bundled non‐pharmacological interventions to no non‐pharmacological treatment or different single or bundled non‐pharmacological interventions in infants receiving opioid treatment for symptoms consistent with opioid withdrawal as defined above.

Non‐pharmacological care is a heterogeneous group of interventions addressing three aspects of care, as follows.

Modifying environmental stimulation: adjustments of environmental stimulation include low light, low noise, clustered care, containment or swaddling, bedding type, positioning, non‐nutritive sucking, bathing, aromatherapy, music therapy, and massage.
Feeding practices: feeding modifications include infant‐led feeding, high caloric feedings, tube feedings, maternal breast milk feedings, non‐maternal breast milk feedings, low‐lactose formula, and breastfeeding.
Support of the mother‐infant dyad: treatments focused on the mother‐infant dyad include parental presence, rooming in, skin‐to‐skin or 'kangaroo care,' parental respite or 'cuddler' programs, trauma‐informed care, social work support, mental health support, treatment program support, and legal support.

Bundled non‐pharmacological interventions: multiple non‐pharmacological interventions delivered in combination.

We assessed non‐pharmacological interventions independently and in combination based on sufficient similarity in population, intervention, and comparison groups studied.

Types of outcome measures

Primary outcomes

Comparison 1

Length of hospitalization (days).
Pharmacological treatment with one or more doses of opioid or sedative medication.

Comparison 2

Length of hospitalization (days).
Length of pharmacological treatment with opioid or sedative medication (days).
Maximum and cumulative dose of opioid medication (milligrams/kilogram and morphine milligram equivalents).

Secondary outcomes

Peak NAS score in first 72 hours of life as measured by a standardized withdrawal assessment tool. Tools included the Finnegan Neonatal Abstinence Scoring Tool (score range 0 to 62) (Finnegan 1975), Lipsitz Neonatal Drug Withdrawal Scoring System (score range 0 to 20) (Lipsitz 1975), Neonatal Narcotic Withdrawal Index (score range 0 to 14) (Green 1981), Neonatal Withdrawal Inventory (score range 0 to 19) (Zahorodny 1998), MOTHER NAS scale (score range 0 to 42) (Jones 2010), or Finnegan Neonatal Abstinence Syndrome Scale – Short Form (score range 0 to 16) (Maguire 2013). Separate assessments were made for each score (Comparison 1 only).
NICU admission (Comparison 1 only).
Length of NICU stay (days).
Physical growth: weight:
1. days to regain birth weight;
2. growth during study period and hospital stay (grams/day); growth velocity (grams/kg/day or change in z‐score);
3. weight nadir (percent weight loss) (Fenton 2017).
Neonatal seizures (any seizures, clinical seizures, electroencephalogram [EEG]‐confirmed seizures; seizure treated with anticonvulsant therapy).
Neonatal and infant all‐cause mortality.
Cerebral palsy (clinical cerebral palsy diagnosed if the child had a non‐progressive motor impairment characterized by abnormal muscle tone and decreased range or control of movements). If data were available, we determined the level of gross motor function using the Gross Motor Function Classification System (Palisano 1997).
Neurodevelopmental outcome at approximately two years' corrected age (acceptable range 18 months to 28 months) including: cerebral palsy, significant mental developmental delay (Bayley Scales of Infant Development Mental Developmental Index less than 70 or Griffith 2 standard deviations [SD] below mean) (Bayley 1993; Chaudhary 2013), legal blindness (less than 20/200 visual acuity), and hearing deficit (aided or less than 60 dB on audiometric testing). The composite outcome 'neurodevelopmental impairment' was defined as having any one of the aforementioned deficits (modified from definitions of moderate‐to‐severe developmental delay) (Schmidt 2007).
Complications of therapy: apnea, need for positive pressure ventilation, need for oxygen, somnolence.
Measures of maternal mood and bonding including Edinburgh Postpartum Depression Scale (Cox 1987), Patient Health Questionnaire (2 or 9) (Kroenke 2001; Whooley 1997), Maternal Attachment Inventory (Muller 1994), Maternal Postpartum Attachment Scale (Condon 1998), Mother‐Infant Bonding Scale (Taylor 2005), or Postpartum Bonding Questionnaire (Brockington 2006). Separate comparisons were made for each score.
Custody status at discharge.
Cost of care: including economic analyses of the impact of care practices including total hospitalization cost or charges per patient.

Post hoc

Peak NAS score in first 72 hours of life definition expanded to include either a 'peak' score at any time or a score in the first 72 hours as measured by a standardized withdrawal assessment tool as described above (Comparison 1 only).

We added a secondary outcome, cost of care, to include economic analyses of the impact of care practices including total hospitalization cost or charges per patient.

Search methods for identification of studies

We used the criteria and standard methods of Cochrane and Cochrane Neonatal (see the Cochrane Neonatal search strategy for specialized register [neonatal.cochrane.org/resources-review-authors]). We searched for errata or retractions from included studies published in full text on PubMed (www.ncbi.nlm.nih.gov/pubmed), and reported the date this was done in the review.

Electronic searches

We conducted a comprehensive search including: Cochrane Central Register of Controlled Trials (CENTRAL 2019, Issue 10) in the Cochrane Library; Ovid MEDLINE(R) and Epub Ahead of Print, In‐Process & Other Non‐Indexed Citations, Daily and Versions(R) (1946 to 11 October 2019); and CINAHL (1981 to 11 October 2019). We included the search strategies for each database in Appendix 1. We applied no language restrictions.

We searched clinical trial registries for ongoing or recently completed trials. We searched the World Health Organization's International Clinical Trials Registry Platform (ICTRP) (www.who.int/ictrp/search/en/), and the US National Library of Medicine's ClinicalTrials.gov (clinicaltrials.gov), via Cochrane CENTRAL. Additionally, we searched the ISRCTN Registry for any unique trials not found through the Cochrane CENTRAL search.

The search strategies for the protocol are in Appendix 2 (Pahl 2018).

Searching other resources

We also searched the reference lists of all identified articles for relevant articles not located in the primary search.

Data collection and analysis

We collected information regarding the method of randomization, blinding, intervention, stratification, and whether the trial was single or multicenter for each included study. We noted information regarding trial participants including gestational age criteria, birth weight criteria, and other inclusion or exclusion criteria. We analyzed the clinical outcomes listed above in Types of outcome measures.

Selection of studies

We included all RCTs, quasi‐RCTs, and cluster trials fulfilling our inclusion criteria. Both superiority trials and non‐inferiority trials were eligible for inclusion. All review authors reviewed the results of the search and separately selected studies for inclusion. We resolved any disagreements by discussion.

We recorded the selection process in sufficient detail to complete a PRISMA flow diagram (Figure 1), and Characteristics of excluded studies table (Moher 2009).

Figure 1

Study flow diagram.

We excluded non‐randomized studies and presented these studies narratively in Appendix 3.

Data extraction and management

Two review authors (AP and RS) extracted, assessed, and coded all data for each study, using a form designed specifically for this review. We replaced any standard error of the mean by the corresponding SD. We resolved any disagreements by discussion. For each study, one review author (AP) entered final data into Review Manager 5 (Review Manager 2020), which a second review author (RS) checked. All review authors reviewed the protocol, analysis, and draft manuscript.

Assessment of risk of bias in included studies

Two review authors (AP and RS) independently assessed the risk of bias (low, high, or unclear) of all included RCTs using the Cochrane 'Risk of bias' tool for the following domains (Higgins 2011).

Sequence generation (selection bias).
Allocation concealment (selection bias).
Blinding of participants and personnel (performance bias).
Blinding of outcome assessment (detection bias).
Incomplete outcome data (attrition bias).
Selective reporting (reporting bias).
Any other bias.

We resolved any disagreements by discussion or by consulting a third review author. See Appendix 4 for a more detailed description of risk of bias for each domain.

Measures of treatment effect

We performed the statistical analyses using Review Manager 5 (Review Manager 2020). We analyzed categorical data using risk ratio (RR) and risk difference (RD). We reported 95% confidence intervals (CIs) for all outcomes. For statistically significant outcomes, we calculated the number needed to treat for an additional beneficial outcome (NNTB) or number needed to treat for an additional harmful outcome (NNTH). We calculated mean differences (MDs) between treatment groups where outcomes were measured in the same way and standardized mean differences (SMDs) where outcomes were measured using different scales for continuous data.

Unit of analysis issues

The unit of analysis was the participating infant in individually randomized trials, and an infant was considered only once in the analysis. The participating neonatal unit or section of a neonatal unit or hospital was the unit of analysis in cluster‐RCTs. While we identified no cluster RCTs, we planned to analyze them using an estimate of the intracluster correlation coefficient (ICC) derived from the trial (if possible), or from a similar trial or from a study with a similar population as described in Section 23.1.4 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019). If we had used ICCs from a similar trial, or from a study with a similar population, we planned to report this and conduct a sensitivity analysis to investigate the effect of variation in the ICC. We identified no cluster trials and performed no sensitivity analysis.

If we identified both cluster‐RCTs and individually randomized trials, we would only combine the results from both if there was little heterogeneity between the study designs, and the interaction between the effect of the intervention and the choice of randomization unit was considered unlikely.

We acknowledged any possible heterogeneity in the randomization unit and performed a sensitivity analysis to investigate possible effects of the randomization unit.

Dealing with missing data

Where feasible, we carried out analysis on an intention‐to‐treat basis for all outcomes. Whenever possible, we analyzed all participants in the treatment group to which they were randomized, regardless of the actual treatment received. If we identified important missing data (in the outcomes) or unclear data, we requested the missing data by contacting the original investigators. We made explicit the assumptions of any methods used to deal with missing data. The search identified few studies for each intervention; therefore, sensitivity analyses were not performed. We addressed the potential impact of missing data on the findings of the review in the Discussion section of the review.

Assessment of heterogeneity

We estimated the treatment effects of individual trials and examined heterogeneity among trials by inspecting the forest plots and quantifying the impact of heterogeneity using the I² statistic. We graded the degree of heterogeneity as: less than 25%, no heterogeneity; 25% to 49%, low heterogeneity; 50% to 75%, moderate heterogeneity; more than 75%, substantial heterogeneity. If we noted statistical heterogeneity (I² greater than 50%), we explored the possible causes (e.g. differences in study quality, participants, intervention regimens, or outcome assessments).

Assessment of reporting biases

We conducted a comprehensive search for eligible studies and were alert for duplication of data. If we had identified 10 or more trials for meta‐analysis, we planned to assess possible publication bias by inspection of a funnel plot. If we uncovered reporting bias that could, in the opinion of the review authors, introduce serious bias, we planned to conduct a sensitivity analysis to determine the effect of including and excluding these studies in the analysis.

Data synthesis

If we identified multiple studies that we considered to be sufficiently similar, we performed meta‐analysis using Review Manager 5 (Review Manager 2020). We determined study similarity based on concordance of population, intervention or combination of interventions, and comparison group. For categorical outcomes, we calculated the typical estimates of RR and RD, each with its 95% CI; for continuous outcomes, we calculated the MD or SMD, each with its 95% CI. We used a fixed‐effect model to combine data where it was reasonable to assume that studies were estimating the same underlying treatment effect. If we judged meta‐analysis to be inappropriate, we analyzed and interpreted individual trials separately. If there was evidence of clinical heterogeneity, we tried to explain this based on the different study characteristics and subgroup analyses.

Subgroup analysis and investigation of heterogeneity

Planned subgroup analyses

Gestational age (late preterm [34 weeks' gestation to 37 weeks' gestation], term (37 weeks' gestation or greater]).
Birth weight (less than 2500 g or 2500 g or greater).
Type of non‐pharmacological intervention:
1. environmental stimulation: adjustments of environmental stimulation include low light, low noise, clustered care, containment or swaddling, bedding type, positioning, non‐nutritive sucking, bathing, aromatherapy, music therapy, and massage;
2. feeding practices: feeding modifications include infant‐led feeding, high‐caloric feedings, tube feedings, maternal breast milk feedings, non‐maternal breast milk feedings, low‐lactose formula, and breastfeeding;
3. support of the mother‐infant dyad: treatments focused on the mother‐infant dyad included parental presence, rooming in, skin‐to‐skin or 'kangaroo care,' parental respite or 'cuddler' programs, trauma‐informed care, social work support, mental health support, treatment program support, and legal support.
Prenatal opioid exposure: polysubstance use, single substance (opioid) use, prescribed opioid, non‐prescribed opioid, not known.
Other prenatal substance exposures:
1. nicotine exposure: nicotine use, no nicotine use;
2. alcohol exposure (based on maternal history or evidence of fetal alcohol spectrum disorder);
3. SSRI exposure.
Maternal social situation (custody status, homelessness, employment status, public insurance, ethnicity, involvement of birth father, marital status, family support).

We did not attempt to perform the subgroup analyses proposed due to insufficient similarity in population, intervention, and comparison groups studied. Few studies were identified for each non‐pharmacological intervention with significant differences between studies.

Sensitivity analysis

We identified few studies for each intervention and did not conduct sensitivity analysis.

Summary of findings and assessment of the certainty of the evidence

We used the GRADE approach, as outlined in the GRADE Handbook (Schünemann 2013), to assess the certainty of evidence for the following (clinically relevant) outcomes for comparison one.

Comparison 1

Length of hospitalization (days).
Pharmacological treatment with opioid or sedative medication.
Peak NAS score in first 72 hours of life.
NICU admission.
Neonatal seizures requiring anticonvulsant therapy.
Neurodevelopmental outcome at approximately two years' corrected age (cerebral palsy, developmental delay [Bayley or Griffith assessment more than two SD below mean] or intellectual impairment (intelligence quotient [IQ] more than two SD below mean), blindness (vision less than 6/60 in both eyes), sensorineural deafness requiring amplification.
Measures of maternal mood and bonding including Edinburgh Postpartum Depression Scale, Patient Health Questionnaire (2 or 9), Maternal Attachment Inventory, Maternal Postpartum Attachment Scale, Mother‐Infant Bonding Scale, or Postpartum Bonding Questionnaire (Brockington 2006; Condon 1998; Cox 1987; Kroenke 2001; Muller 1994; Taylor 2005; Whooley 1997)

We planned to assess the certainty of evidence for the following (clinically relevant) outcomes for comparison two.

Comparison 2

Length of hospitalization (days).
Length of pharmacological treatment with opioid or sedative medication (days).
Maximum and cumulative dose of opioid medication (morphine milligram equivalents).
Length of NICU stay.
Neonatal seizures requiring anticonvulsant therapy.
Neurodevelopmental outcome at approximately two years' corrected age (cerebral palsy, developmental delay [Bayley or Griffith assessment more than two SD below mean] or intellectual impairment [IQ] more than two SD below mean), blindness (vision less than 6/60 in both eyes), sensorineural deafness requiring amplification.
Measures of maternal mood and bonding including Edinburgh Postpartum Depression Scale, Patient Health Questionnaire (2 or 9), Maternal Attachment Inventory, Maternal Postpartum Attachment Scale, Mother‐Infant Bonding Scale, or Postpartum Bonding Questionnaire (Brockington 2006; Condon 1998; Cox 1987; Kroenke 2001; Muller 1994; Taylor 2005; Whooley 1997).

Two review authors (AP and RS) independently assessed the certainty of the evidence for each of the outcomes listed above. We considered evidence from RCTs as high certainty, downgrading the evidence one level for serious (or two levels for very serious) limitations based upon the following: design (risk of bias), consistency across studies, directness of the evidence, precision of estimates, and presence of publication bias.

We used GRADEpro GDT to create summary of findings Table 1 to report the certainty of the evidence for comparison one.

The GRADE approach results in an assessment of the certainty of a body of evidence as one of four grades.

High certainty: further research is very unlikely to change our confidence in the estimate of effect.
Moderate certainty: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low certainty: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low certainty: we are very uncertain about the estimate.

Results

Description of studies

Included studies

The search identified six RCTs evaluating non‐pharmacological care for opioid withdrawal published between 1975 and 2018 (Characteristics of included studies table). Four studies compared modifications of environmental stimulation (D'Apolito 1999; Maichuk 1999; Oro 1988; Ostrea 1975), one study compared feeding practices (Bogen 2018), and one study compared interventions targeting support of the mother‐infant dyad (MacVicar 2018). Two studies were RCTs due to use of computer‐generated randomization (Bogen 2018; MacVicar 2018), one was classified as a quasi‐RCT due to randomization by hospital record number (Oro 1988), and three reported randomization of study participants without further details about the methods and may represent RCTs or quasi‐RCTs (D'Apolito 1999; Maichuk 1999; Ostrea 1975). The search identified no cluster trials.

Excluded studies

The search identified 34 non‐randomized studies published between 2005 and 2018 comparing single or bundled non‐pharmacological interventions to no non‐pharmacological treatment or different single or bundled non‐pharmacological interventions for opioid withdrawal in newborns (see Characteristics of excluded studies table; Table 1; Table 2; Table 3; Table 4; and Appendix 3).

Table 1. Excluded studies of modifying environmental stimulation

Study	Details
Radziewicz 2018	Methods	Prospective cohort
	Participants	30 newborns at risk for withdrawal in a newborn nursery
	Time period	Not reported
	Site of care	Newborn nursery (Ohio, USA)
	Interventions	Quote: "Each subject received one thirty‐minute treatment of Reiki in a private room."
	Outcomes	Slight decrease in heart rate and withdrawal scores following treatment
	Notes	—
Zuzarte 2017	Methods	Prospective cohort (within‐subjects)
	Participants	26 opioid‐exposed newborns (> 37 weeks' gestation) treated pharmacologically for NAS
	Time period	Not reported
	Site of care	Neonatal intensive care unit and newborn nursery (Massachusetts, USA)
	Interventions	Quote: "A specially‐constructed mattress delivered low‐level SVS (30–60Hz, 10–12μm RMS [root‐mean‐square]), alternated in 30‐min intervals between continuous vibration (ON) and no vibration (OFF) over a 6–8 hr session."
	Outcomes	35% reduction in movement activity with SVS
	Notes	—

NAS: neonatal abstinence syndrome; SVS: stochastic vibrotactile stimulation.

Table 2. Excluded studies of feeding practices

Study	Details
Abdel‐Latif 2006	Methods	Retrospective cohort
	Participants	190 consecutive drug‐dependent mother‐infant pairs
	Time period	1998–2004
	Site of care	Postnatal ward with mother unless medical or social contraindications (New South Wales, Australia)
	Interventions	Predominant feed on day 5 of life: ≤ 2 feeds daily of formula (85 infants) or > 2 feeds daily of formula (105 infants)
	Outcomes	≤ 2 feeds daily of formula vs > 2 feeds daily of formula Shorter length of hospitalization (mean: 14.7 days, SD 14.9 vs 19.1 days, SD 15.0; P = 0.049) Significant decrease in need for treatment (52.9% vs 79.0%; P < 0.001) Non‐significant decrease in treatment duration (mean: 85.4 days, SD 71.7 vs 108.2 days, SD 81.8; P = 0.185) Non‐significant decrease in maximum dose of morphine (mean: 0.57 mg/kg/day, SD 0.22 vs 0.59 mg/kg/day, SD 0.22; P = 0.526) Fewer infants in foster care (9.4% vs 29.5%; P = 0.001) Fewer infants designated; quote: "child at risk" (31.8% vs 68.6%; P < 0.001) Lower mean Finnegan scores (numeric values not provided in text) Later median time to withdrawal (10 days vs 3 days; P < 0.001)
	Notes	More heroin, non‐opioid, and polydrug use in formula group Emailed authors to clarify data
Arlettaz 2005	Methods	Retrospective cohort
	Participants	86 infants of women enrolled in a methadone maintenance program
	Time period	1996–2001
	Site of care	Unit not specified, 77% transferred to the neonatal unit for medical reasons or pharmacological treatment (Switzerland)
	Interventions	Infants who received more than half of total amount of milk during hospital stay as human milk (27 infants) vs infants who received less than half (59 infants)
	Outcomes	Quote: "Twenty‐six percent of breast‐fed babies (7/27) and 78% of formula‐fed babies (42/54) developed NAS (P < 0.01)." Pharmacological treatment rate and custody not reported by feeding status
	Notes	—
Dryden 2009	Methods	Retrospective cohort
	Participants	450 singleton pregnancies of drug misusing women prescribed substitute methadone for pregnancy 354 term substitute methadone‐exposed infants included in subgroup published in 2012
	Time period	January 2004 to December 2006
	Site of care	Postnatal ward with mother unless any specific indication for admission to the neonatal unit (UK)
	Interventions	Received maternal breast milk ≥ 72 hours (99 infants) vs formula fed or received maternal breast milk < 72 hours (351 infants)
	Outcomes	Received maternal breast milk ≥ 72 hours vs formula fed or received maternal breast milk < 72 hours Breastfeeding associated with reduced odds of requiring treatment for NAS (OR 0.55, 95% CI 0.34 to 0.88; P = 0.013) Follow‐up analysis of 354 infants with complete data (breastfed 72 infants, non‐breastfed 282 infants) showed higher median maximal weight loss in breastfed group (10.2% vs 8.5%; P = 0.003) The group of 22 exclusively breastfed infants had a median maximal weight loss of 9.3% Quote: "For infants who were not admitted to the [neonatal unit] (n=208), median age at discharge was 7 days; this did not differ between breast and formula feeders" More breastfed infants were classified as "non‐admitted" and did not require admission to the neonatal unit (52/72 [72%] vs 155/282 [55%])
	Notes	Emailed authors to clarify the relationship between the 2 publications, 2009 study is the primary study as fewer infants were included in 2012 study due to missing data.
Hodgson 2012	Methods	Retrospective cohort
	Participants	295 singleton mother‐infant dyads who received care in combined care for maternal substance use in a rooming in program
	Time period	October 2003 to December 2006
	Site of care	Postnatal; quote: "combined care nursery" with mother including those receiving pharmacotherapy (British Columbia, Canada)
	Interventions	Breastfed (196 infants) vs not breastfed (99 infants)
	Outcomes	Significant decrease in probability of an infant receiving morphine (12% of breastfed infants vs 37% of non‐breastfed infants; OR 0.21, SD 1.43, 95% CI 0.10 to 0.42; P < 0.001)
	Notes	—
Isemann 2011	Methods	Retrospective cohort
	Participants	128 infants that received pharmacotherapy for opiate withdrawal in the newborn intensive care unit (92 infants > 37 weeks, 36 infants < 37 weeks)
	Time period	January 2002 to December 2007
	Site of care	NICU (Ohio, USA)
	Interventions	Mother's breast milk ingested vs formula fed
	Outcomes	Inverse correlation between the percentage of mother's breast milk ingested and length of stay (β = –0.03; P = 0.02) Median: 12.5 days, range 3–51 days for infants who ingested maternal breast milk vs 18.5 days, range 9̫–43 days for formula‐fed infant
	Notes	—
Jansson 2008	Methods	Prospective case control
	Participants	8 breastfed methadone‐exposed infants and 8 matched formula‐feeding infants
	Time period	January 2001 to September 2005
	Site of care	Unit not specified (Maryland, USA)
	Interventions	Breastfed (8 infants) vs formula fed (8 infants)
	Outcomes	Breastfed vs formula fed Decrease in use of pharmacological treatment (12.5% vs 50%; P = 0.28) Shorter time of treatment (median: 6 days, IQR not reported vs 13.5 days, IQR 10.5) Lower day 3 NAS score (median: 4.5, IQR 3.9 vs 6.8, IQR 1.2; P non‐significant) Lower peak NAS score (median: 6.5, IQR 4.5 vs 11.0, IQR 5.0; P non‐significant)
	Notes	More medical comorbidities in breastfed group 3 breastfeeding infants not included due to failure to provide complete study specimens 1 breastfeeding infants not included due to inability to find match
Liu 2015	Methods	Retrospective cohort
	Participants	194 methadone‐maintained mother‐infant dyads > 30 weeks' gestation
	Time period	2000–2006
	Site of care	Unit not specified; quote: "two birthing units" (New South Wales, Australia)
	Interventions	Categorized in the 1st 2 days of life as predominantly breastfed (≥ 50% of feeds, 32 infants), fed expressed breast milk (≥ 15 mL expressed breast milk and breastfed < 3 times daily, 12 infants), or formula fed (≥ 50% of feeds and < 15 ml/day breast milk ingestion, 150 infants)
	Outcomes	Breastfed vs fed expressed breast milk vs formula fed No difference in percent requiring treatment for NAS (72% vs 100% vs 81%; P = 0.11) No difference in maximum dose of morphine required (median: 0.5 mg/kg/day, IQR 0.5 to 0.7 vs 0.7, IQR 0.53 to 9.84 vs 0.5, IQR 0.5 to 0.7; P = 0.20) No difference in mean NAS score (5.1, SD 1.3 vs 5.7, SD 0.9 vs 5.4, SD 1.1; P = 0.47) Quote: "Breastfeeding during the first 2 days of life was associated with a delayed onset of NAS (P = .04)"
	Notes	—
McQueen 2011	Methods	Retrospective cohort
	Participants	28 term infants exposed to methadone with symptoms of NAS
	Time period	March 2007 to March 2008
	Site of care	Unit not specified (Ontario, Canada)
	Interventions	Breastfeeding (> 75% breast milk, 8 infants) vs combination feeds (25% to 75% breast milk, 11 infants) vs formula fed (< 25% breast milk, 9 infants)
	Outcomes	Breastfed vs combination feed vs formula fed Fewer mean NAS scores on Modified Finnegan Scoring Tool (25, SD 23.5 vs 56.2, SD 39.1 vs 95.6, SD 34.6) Lower mean NAS scores on Modified Finnegan Scoring Tool (4.9, SD 2.9 vs 6.5, SD 3.7 vs 6.9, SD 4.2) Fewer infants scored > 8 (17% vs 38.8% vs 40.2%)
	Notes	More regular prenatal care in the breastfed group Emailed with authors to clarify data
Metz 2015	Methods	Retrospective cohort
	Participants	390 infants born to opioid‐dependent women undergoing comprehensive treatment during pregnancy (184 methadone, 77 buprenorphine, 129 slow‐release oral morphine) 50 infants excluded due to incomplete data
	Time period	1994 to January 2009
	Site of care	Unit not specified (Austria)
	Interventions	Breastfed infants (100 infants; methadone group 48 infants, slow‐release oral morphine group 21 infants, buprenorphine group 31 infants) vs non‐breastfed infants (250 infants; methadone group 118 infants, slow‐release oral morphine group 91 infants, buprenorphine group 41 infants)
	Outcomes	Breastfed vs non‐breastfed Decreased mean modified Finnegan score in methadone group (6.52, SD 2.50 vs 7.94, SD 2.87; P = 0.004) and slow‐release oral morphine group (6.97, SD 2.77 vs 9.07, SD 3.11; P = 0.007) with no difference in buprenorphine group (4.89, SD 2.03 vs 5.77, SD 2.49; P = 0.110) Decreased peak modified Finnegan score in methadone group (mean: 12.60, SD 5.16 vs 17.00, SD 5.65; P < 0.001) and slow‐release oral morphine group (mean 15.24, SD 6.66 vs 18.93, SD 5.56; P = 0.009) with no difference in buprenorphine group (mean: 11.19, SD 5.11 vs 12.29, SD 4.78; P = 0.352) Decreased total morphine dose in methadone group (mean: 4.35 mg, SD 9.03 vs 12.65 mg, SD 18.64; P < 0.001) with no difference in slow‐release oral morphine group (mean: 5.23 mg, SD 7.00 vs 8.75 mg, SD 11.54; P = 0.193) and buprenorphine group (mean: 1.90 mg, SD 3.32 vs 2.77 mg, SD 5.78; P = 0.457) Decreased duration of treatment in methadone group (mean: 8.10 days, SD 12.64 vs 16.94 days, SD 16.75; P < 0.001) and slow‐release oral morphine group (mean: 10.20 days, SD 12.53 vs 18.07, SD 13.67; P = 0.020) with no difference in buprenorphine group (mean: 4.74 days, SD 5.58 vs 5.76 days, SD 7.37; P = 0.524) Decreased length of stay in methadone group (mean: 17.17 days, SD 10.28 vs 29.36 days, SD 15.07; P < 0.001) and slow‐release oral morphine group (mean: 19.71 days, SD 12.37 vs 31.03 days, SD 13.81; P = 0.001) with no difference in buprenorphine group (mean: 12.26 days, SD 5.35 vs 14.71 days, SD 9.60; P = 0.174)
	Notes	Outcomes reported as subgroups only
O'Connor 2013	Methods	Retrospective cohort
	Participants	85 mother‐infant dyads in buprenorphine treatment program, infants > 35 weeks
	Time period	December 2007 to August 2012
	Site of care	Unit not specified (Maine, USA)
	Interventions	Breastfeeding (65 infants) vs non‐breastfed infants (20 infants)
	Outcomes	Breastfed vs non‐breastfed Non‐significant increase in length of hospital stay (mean: 7.08 days, SD 4.39 vs 6.60 days, SD 1.70; P = 0.35) Non‐significant decrease in peak modified Finnegan NAS score (mean: 8.83, SD 3.56 vs 9.65, SD 2.58; P = 0.17) Non‐significant decrease in pharmacological treatment (23.1% vs 30.0%; P = 0.56) Non‐significant decrease in time to first and last peak modified Finnegan NAS score (first: mean: 66.5 hours, SD 43.80 vs 73.5 hours, SD 41.82; P =.32; last: mean: 76.1 hours, SD 43.95 vs 78.3 hours, SD 38.54; P = 0.67) Decreased likelihood of higher modified Finnegan NAS scores (≥ 8: 65% vs 75%; ≥ 12: 17% vs 30%)
	Notes	Breastfed infants were significantly less likely to have exposure to illicit substances in the 3rd trimester (41.4% vs 80.0%; P = 0.004)
Pritham 2012	Methods	Retrospective cohort
	Participants	152 opioid‐dependent pregnant women in treatment (136 methadone, 16 buprenorphine) and their infants
	Time period	January 2005 to December 2007
	Site of Care	NICU (Maine, USA)
	Interventions	Breastfed (17 infants, methadone group 14, buprenorphine group 3) vs both formula and breast (26 infants, methadone group 22, buprenorphine group 4) vs formula fed (105 infants, methadone group 96, buprenorphine group 9)
	Outcomes	Length of stay was shorter in breastfed neonates than formula‐fed neonates or neonates that fed both formula and breast in regression model of methadone group (β = –0.176; P = 0.05)
	Notes	—
Radmacher 2017	Methods	Prospective cohort (historically controlled feasibility study)
	Participants	12 term formula‐fed infants receiving oral morphine for opioid withdrawal matched to historical controls
	Time period	2014–2017
	Site of care	NICU (Kentucky, USA)
	Interventions	Exclusive donor milk for up to 2 weeks (12 infants) vs formula‐fed historical controls (12 infants)
	Outcomes	Donor milk vs formula fed No significant difference in length of stay (mean: 41 days, SD 16 vs 33 days, SD 9; P non‐significant) No significant difference in length of treatment (mean: 37 days, SD 16 vs 29 days, SD 9; P non‐significant) No significant difference in median total Finnegan Neonatal Abstinence Score (values not reported) Significant decrease in gastrointestinal subscore > 2 (39% vs 61%; P = 0.001) No significant difference in return to birth weight (mean: 17 days, SD 7 vs 16 days, SD 4; P non‐significant) No significant difference in weight gain during study period (mean: 19.1 g/day, SD 13.1 vs 20.5 g/day, SD 3.0; P non‐significant)
	Notes	Benzodiazepine exposure in the donor milk group but not the control group Significant deviation from intended intervention noted
Short 2016	Methods	Retrospective cohort
	Participants	3725 singleton in‐hospital births > 35 weeks' gestation diagnosed with NAS based on discharge diagnosis codes
	Time period	2012–2014
	Site of care	Unit not specified, though mentions benefits of rooming in and 40% admission to NICU (Pennsylvania, USA)
	Interventions	Breastfeeding at discharge (1576 infants) vs not breastfeeding (1968 infants) as reported on birth certificate
	Outcomes	Length of stay was 9.4% shorter in infants breastfed at discharge than infants not breastfed at discharge in regression model (β = –0.060; P = 005)
	Notes	No information about in‐utero exposure or postnatal treatment for NAS Breastfed infants were more likely to have a higher birthweight and gestational age; less likely to be insured by Medicaid; and more likely to be born to mothers who completed high school, were married, had adequate prenatal care, and who did not smoke
Wachman 2013	Methods	Prospective cohort
	Participants	86 infants > 36 weeks' gestation exposed to methadone or buprenorphine in utero
	Time period	July 2011 to July 2012
	Site of care	Unit not specified, 5 tertiary care centers and community hospitals (Massachusetts and Maine, USA)
	Interventions	Assessed breastfeeding as a potential covariate related to NAS severity; breastfeeding (38 infants) vs no breastfeeding (48 infants)
	Outcomes	Breastfed vs non‐breastfed Shorter length of stay (mean: 15.8 days, 95% CI 11.5 to 20.1 vs 27.4 days, 95% CI 22.5 to 32.3; P < 0.001) Decrease in pharmacological treatment (50% vs 77%; P = 0.009)
	Notes	Objective: to determine whether single nucleotide polymorphisms in the μ‐opioid receptor (OPRM1), multiple‐drug resistance (ABCB1), and catechol‐O‐methyltransferase (COMT) genes are associated with length of hospital stay and the need for treatment of NAS
Welle‐Strand 2013	Methods	Retrospective and prospective cohort
	Participants	124 infants of women on opioid maintenance therapy (78 methadone, 46 muprenorphine)
	Time period	1999–2003 (36 infants, retrospective), 2005–2007 (36 infants, prospective), 2007–2009 (52 infants, retrospective)
	Site of care	Unit not specified, national cohort (Norway)
	Interventions	Breastfed (95 infants; methadone exposed 58, buprenorphine exposed 37) vs not breastfed (29 infants; methadone exposed 20, buprenorphine exposed 9)
	Outcomes	Breastfed vs not breastfed Non‐significant decrease in NAS use of pharmacological treatment overall (overall: 57% vs 69%, not significant; methadone exposed: 53% vs 80%; P < 0.05; buprenorphine exposed: 64% vs 44%; P not significant) Significant decrease in NAS treatment duration overall (overall: mean: 28.6 days, SD 19.1 vs 46.7 days, SD 26.3; P < 0.05; methadone‐exposed mean: 31.0 days, SD 21.4 vs 48.9 days, SD 27.2; P < 0.05; buprenorphine‐exposed mean: 25.7 days, SD 16.0 vs 38.8 days, SD 24.0; P not significant) Length of NAS treatment was shorter in breastfed infants vs not breastfed infants in regression model (overall: β = –19.7; P = 0.00; methadone exposed: β = –20.7; P = 0.00; buprenorphine exposed: β = –6.62; P = 0.59)
	Notes	More preterm birth and delivery by cesarean in non‐breastfed group

CI: confidence interval; NAS: neonatal abstinence syndrome; IQR: interquartile range; NICU: neonatal intensive care unit; OR: odds ratio; SD: standard deviation.

Table 3. Excluded studies of support of the mother‐infant dyad

Study	Details
Abrahams 2007	Methods	Retrospective cohort
	Participants	106 infants of mothers with known heroin or methadone use during pregnancy
	Time period	October 2001 to December 2002
	Site of care	Postnatal ward and intensive care nursery (BC, Canada)
	Interventions	Rooming in infants in study hospital (32 infants) compared to historical comparisons in study hospital (38 infants) and comparisons at a neighboring hospital without rooming in (36 infants)
	Outcomes	Rooming in vs historical comparisons in study hospital vs neighboring hospital without rooming in Shorter length of hospital stay (mean: 11.8 days, SD 9.1 vs 23.5 days, SD 24.6 vs 25.9 days, SD 19.7) Statistically significant decrease in use of pharmacological treatment (25.0% vs 55.3% [RR 0.45, 95% CI 0.23 to 0.87] vs 52.8% [RR 0.47, 95% CI 0.24 to 0.93]) Statistically significant decrease in pharmacological treatment duration (mean: 5.9 days, SD 14.2 in study hospital vs 18.6 days, SD 23.4 vs 18.6 days, SD 20.1) Statistically significant increase in likelihood of discharge with mothers (discharged in custody of mothers: 71.9% vs 31.6% [RR 2.28, 95% CI 1.36 to 3.81] vs 42.5% [RR 1.52, 95% CI 1.01 to 2.29]) Trend toward decreased prevalence of NAS symptoms Increased prevalence of weight loss > 10% in the first week of life (16.6% vs 5.2% [RR 2.96, 95% CI 0.62 to 14.28] vs 5.3% [RR 2.81, 95% CI 0.59 to 13.5]) Statistically significant decrease in admission to a level II nursery (37.5% vs 89.5% [RR 0.42, 95% CI 0.26 to 0.66] and 83.3% ]RR 0.45, 95% CI 0.28 to 0.72])
	Notes	Several baseline differences between groups and hospitals with significant difference in feeding practices between groups (62.5% of the rooming in group breastfeeding compared to 7.9% of the historical cohort and 11.1% of the concurrent cohort)
Abrahams 2010	Methods	Retrospective cohort
	Participants	952 singleton infants of mothers with documented substance use
	Time period	October 2003 to December 2006
	Site of care	Postnatal; quote: "combined care nursery" and intensive care nursery (British Columbia, Canada)
	Interventions	Rooming in with comprehensive education in study hospital (355 infants, term 255 infants) vs standard nursery care at all other hospitals (597 infants, term 434 infants)
	Outcomes	Rooming in vs standard nursery care Longer hospital length of stay (all infants: mean: 20.7 days, SD 20.5 vs 10.7 days, SD 19.4; term infants: mean 17.6 days, SD 11.3 vs 8.1 days, SD 12.6) Decreased admissions to NICU (all infants: 38.9% vs 45.0%; OR 0.68, 95% CI 0.51 to 0.92; P < 0.001; term infants: 23.5% vs 38.5%; OR 47, 95% CI 0.324 to 0.67; P < 0.001) Shorter NICU length of stay in term infants (all infants: mean: 5.0 days, SD 19.2 vs 5.0 days, SD 12.6; term infants: mean: 1.1 days, SD 3.1 vs 3.1 days, SD 8.3) Increased receipt of breast milk (all infants: 63.7% vs 45.4%; OR 2.11, 95% CI 1.61 to 2.77; P < 0.001) No difference in breastfeeding status at discharge (all infants: 34% vs 35.1%; OR 0.953, 95% CI 0.67 to 1.35; P = 0.79) Increased discharge home with mother (69.9% vs 58.7%)
	Notes	Baseline differences between groups and hospitals. Newborns in rooming in group had increased odds of receiving breast milk during hospital stay.
Crook 2017	Methods	Retrospective and prospective cohort (quality improvement methodology)
	Participants	200 infants evaluated for NAS
	Time period	February 2014 to July 2015
	Site of care	NICU, small percentage on pediatric unit if high NICU census (North Carolina, USA)
	Interventions	Baby friendly status Breastfeeding initiative (3 class curriculum)
	Outcomes	Decrease in length of stay from a mean of 18.80 days, SD 14.71 to 10.41 days, SD 10.54 (P < 0.001) Decrease in pharmacological treatment from 67.3% to 34.8% (P < 0.001) Increase in exclusive breastfeeding from 20.0% to 31.9% (P not reported)
	Notes	Quote: "Infants who received more than 50% human milk had a predicted [length of stay] 2.78 days shorter than infants who received more than 50% formula during hospitalization (P = .08)."
Howard 2017	Methods	Retrospective cohort
	Participants	86 mother‐infant dyads in a single center rooming in cohort of infants treated pharmacologically for NAS
	Time period	March 2015 to April 2016
	Site of care	Pediatric inpatient unit for pharmacological treatment (Massachusetts, USA)
	Interventions	Parental presence (percent of time)
	Outcomes	Maximum (100%) parental presence was associated with a 9‐day shorter length of stay (r = –0.31, 95% CI –0.48 to –0.10; P = 0.01), 8 fewer days of infant opioid therapy (r = –0.34, 95% CI –0.52 to –0.15; P = 0.001), and 1 point lower mean Finnegan score (r = –0.35, 95% CI –0.52 to –0.15; P = 0.01) Quote: "There was a 5.3 mg decrease in the total morphine equivalent dose with increased parental presence (r = –0.20; 95% CI, –0.39 to 0.02; P = .06)" Quote: "The mean NAS score when a parent was present was significantly lower compared with when a parent was not present (5.1 [95% CI, 4.9–5.3] versus 6.0 [95% CI, 5.8–6.2]; P < .0001)" Quote: "Parental presence was higher for infants who were breastfed (65.2% vs 44.5%; P < .001)" Quote: "Parental presence was lower for infants in Department of Children and Families (DCF) custody (36.6% vs 59.7%; P < .001)" Quote: "After adjusting for breastfeeding, parental presence remained significantly associated with reduced NAS score and opioid treatment days"
	Notes	Quote: "In unadjusted analyses, any amount of breastfeeding was associated with a decreased [length of stay] (16.5 vs 21.2 days; P < .01) as was clinical trial participation (15.8 vs 20.0 days; P < .01). Similarly, breastfeeding and clinical trial participation were also significantly associated with decreased duration of opioid therapy and decreased total morphine equivalent dose."
Hünseler 2013	Methods	Retrospective cohort
	Participants	77 infants with NAS transferred from the maternity ward to the neonatal ward for potential pharmacological treatment
	Time period	2004–2011
	Site of care	Neonatal ward (Germany)
	Interventions	Rooming in within the neonatal ward (24 infants) vs not rooming in within the neonatal ward (53 infants)
	Outcomes	Rooming in vs no rooming in Shorter length of hospital stay (mean: 36.6, SD 10.2 vs 42.8, SD 15.3; P = 0.077) No difference in use of pharmacological treatment (79% vs 88.7%; P = 0.14) Shorter duration of pharmacological treatment (mean: 29.5, SD 11.5 vs 35.8, SD 15.3; P = 0.043) Lower maximum dose of opioid (mean: 132 μg/kg/day diluted tincture of opium, SD 56 vs 145 μg/kg/day diluted tincture of opium, SD 60; P = 0.84) No difference in the frequency of Finnegan Scores ≥ 12 (6.4% vs 6.3%; P = 0.98). No difference in maximum Finnegan score (19 points in both groups). Lower costs (mean EUR 10,620, SD 5306 vs EUR 14,331, SD 6028; P = 0.014) More infants discharged home with families (79.2% vs 69.8%; P = 0.046)
	Notes	Only 2 infants were breastfed
McKnight 2016	Methods	Retrospective cohort (historical controls)
	Participants	44 singleton infants > 36 weeks' gestation of women taking chronic opioid therapy
	Time period	May 2012 to September 2014 (excluding June to August 2013)
	Site of care	Pediatrics unit with transfer to NICU for pharmacological treatment (Ontario, Canada)
	Interventions	Baseline (24 infants) and postimplementation (20 infants) of initial observation while rooming in with mother in a private room in the pediatrics unit. Infants transferred to the NICU if pharmacological treatment was required
	Outcomes	Shorter length of hospitalization (median: 5.0, IQR 3 vs 24.0, IQR 34; P < 0.001) Reduced use of pharmacological treatment of NAS (15.0% vs 83.3%; P < 0.001) No difference in duration of treatment with morphine (median: 24.0, IQR could not be calculated vs 29.5, IQR 23; P = 0.83) No difference in maximum daily dose of morphine (median: 0.48 mg/kg/day, IQR could not be calculated vs 0.48 mg/kg/day, IQR 0.48; P = 0.90) Fewer treated with adjunct pharmacotherapy (0% vs 30.0%; P = 0.54)
	Notes	Quote: "While not significant, a higher proportion of infants in the rooming‐in group were breastfed exclusively or in combination with formula." Emailed authors to clarify data and the relationship between the 2 publications, 1 additional participant included in earlier publication (Newman 2015) who was different from the usual population
Metz 2011	Methods	Retrospective cohort
	Participants	114 opioid‐exposed newborns of mothers maintained on methadone (70 infants) or buprenorphine (44 infants)
	Time period	June 2005 to December 2009
	Site of care	Unit not specified (Austria)
	Interventions	Participation in a randomized double‐blind double‐dummy clinical trial ("MOTHER study", 37 infants) vs standard care (77 infants) Clinical trial included daily maternal contact with study staff and mandatory rooming in with mother for ≥ 10 days after delivery Standard care neonates were treated in a separate ward and mothers were discharged when medically cleared
	Outcomes	Clinical trial vs standard protocol Shorter length of stay (days) in the clinical trial group (methadone‐exposed mean: 16.74, SD 6.78 vs 29.36, SD 17.94; buprenorphine‐exposed mean: 13.67, SD 2.63 vs 13.92, SD 7.33) More infants used pharmacological treatment in the clinical trial group (76% vs 60%) Lower total medication dose in the clinical trial group (methadone‐exposed mean: 5.18 mg, SD 6.94 vs 21.61 mg, SD 26.64; buprenorphine‐exposed mean: 2.02 mg, SD 1.71 vs 4.30 mg, SD 7.30) Shorter length of pharmacological treatment in the clinical trial group (methadone‐exposed mean: 9.53 days, SD 9.14 vs 21.25 days, SD 21.22; buprenorphine‐exposed mean: 7.33 days, SD 4.46 vs 6.62 days, SD 8.07)
	Notes	Modified Finnegan used for clinical trial vs traditional Finnegan for standard care Likely fundamentally different group self‐selected for clinical trial
Ordean 2015	Methods	Retrospective cohort
	Participants	94 pregnant methadone‐maintained women attending integrated care programs and their children
	Time period	1997–2009
	Site of care	Unit not specified for rooming in cohort, NICU (Ontario, British Columbia, and Quebec, Canada)
	Interventions	Rooming in approach (36 infants) vs 2 neighboring hospitals that admit infants to the neonatal intensive care unit for NAS scoring (36 and 22 infants)
	Outcomes	Rooming in vs hospital 1 (no rooming in) vs hospital 2 (no rooming in) Longer hospital stay (mean: 26 days, SD 16 vs 14 days, SD 15 vs 16 days, SD 14) No difference in use of pharmacological treatment (29% vs 31% vs 18%; P = 0.564) No difference in duration of NAS treatment (mean: 13 days, SD 6 vs 25 days, SD 17 vs 30 days, SD 27; P = 0.086) Decreased NICU admission (42% vs 94% vs 91%; P = 0.003) No difference in weight loss (7.4% vs 7.0% vs 8.7%; P = 0.204) No difference in days to weight gain (mean: 4.3, SD 1.9 vs 4.2, SD 2.5 vs 5.6, SD 2.8; P = 0.093) Increase in discharge with mother (41% vs 69% vs 68%; P = 0.115) No difference in breastfeeding (14% vs 17% vs 22%; P = 0.767)
	Notes	Minimal difference in breastfeeding between groups. Significant intersite differences were found in ethnicity and marital status. Emailed authors to clarify data.
Saiki 2010	Methods	Retrospective cohort (historical controls)
	Participants	60 infants with NAS
	Time Period	2002–2007
	Site of Care	Postnatal ward for monitoring and pharmacological treatment and neonatal unit (UK)
	Interventions	Baseline (42 infants) and postimplementation (18 infants) of a system of pharmacological treatment of the infant in the postnatal ward instead of transferring from the postnatal ward to the neonatal unit if escalating symptoms
	Outcomes	Baseline vs postimplementation Shorter hospital stay (mean: 19.8 days vs 15.9 days; P = 0.012) Fewer infants required treatment (45% vs 11%; P = 0.012) Shorter duration of treatment (mean: 12.7 days vs 7.3 days; P = 0.05) No difference in discharge with mother (60% vs 67%; P = 0.264)
	Notes	—
Summey 2018	Methods	Retrospective cohort
	Participants	466 infants receiving care for NAS
	Time period	2006–2014
	Site of care	Postnatal ward for monitoring and pharmacological treatment and Neonatal unit (South Carolina, USA)
	Interventions	Infants enrolled in the Managing Abstinence in Newborns (MAiN) intervention (110 infants) vs infants who were potentially eligible but did not receive the intervention (356 infants) MAiN; quote: "provides multidisciplinary, coordinated, community‐based care for infants at risk for opioid withdrawal and their families within a mother/baby unit. MAiN's three essential elements are early treatment (methadone administered within 48 hours of birth), the option to room in with the mother for the entire hospitalization, and a combined inpatient/outpatient wean."
	Outcomes	MAiN vs no MAiN Non‐significant decrease in length of stay (median: 8 days, IQR 6 to 10 vs 9 days, IQR 4 to 18; P = 0.14) Significantly higher involvement of child protective services in the MAiN cohort (32.7% vs 23.3%; P = 0.048) All infants received level 1 care, 68.8% of comparison babies received level 2 to 3 care Lower total cost (median: USD 10,058, IQR USD 7935 to USD 11,518 vs USD 18,262, IQR USD 5816 to USD 40,922; P < 0.001) 'Convulsions' were reported in 1 intervention infant and no comparison infants Diagnosed with 'abnormal loss of weight' more often (3.6% vs 0.8%; P = 0.04)
	Notes	No significant difference in breastfeeding between groups (43.6% vs 48.9%; P = 0.57)

CI: confidence interval; DCF: Department of Children and Families; IQR: interquartile range; MAiN: Managing Abstinence in Newborns; MOTHER: Maternal Opioid Treatment: Human Experimental Research; NAS: neonatal abstinence syndrome; NICU: neonatal intensive care unit; OR: odds ratio; r: Pearson's correlation coefficient; RR: risk ratio; SD: standard deviation.

See: Summary of findings 1 Non‐pharmacological treatment compared to placebo for infants at risk for, or having early symptoms consistent with, opioid withdrawal

Table 4. Excluded studies of multiple non‐pharmacological care practices

Study	Details
Grossman 2017	Methods	Retrospective and prospective cohort (quality improvement methodology)
	Participants	287 infants diagnosed with NAS. Excluded infants not exposed to methadone and with serious comorbid conditions
	Time period	January 2008 to June 2016
	Site of care	Pediatrics unit, postnatal ward, and NICU (Connecticut, USA)
	Interventions	Quote: "Standardization of nonpharmacologic care coupled with an empowering message to parents, development of a novel approach to assessment, administration of morphine on an as‐needed basis, and transfer of infants directly to the inpatient unit." Baseline period from January 2008 to February 2010 (55 infants), intervention period (188 infants), and postimplementation period from May 2015 to June 2016 (44 infants)
	Outcomes	Decrease in mean length of stay (22.4 days to 5.9 days; P < 0.001), postnatal morphine treatment (98% to 14%; P < 0.001), and mean cost (USD 44,824 to USD 10,289; P < 0.001) More infants fed breast milk at discharge following implementation (20% to 45%; P = 0.01) Fewer infants admitted to the NICU (100% to 20%; P < 0.001)
	Notes	—
Holmes 2016	Methods	Retrospective and prospective cohort (quality improvement methodology)
	Participants	163 opioid‐exposed newborns Excluded infants with gestational age < 35 weeks, another reason for NICU admission, or who completed treatment at another facility
	Time period	March 2012 to February 2015
	Site of care	Pediatrics unit, postnatal ward, and NICU (New Hampshire, USA)
	Interventions	Staff training, family education (involvement in symptom monitoring and non‐pharmacological treatment), and care in inpatient pediatric unit Baseline period from March 2012 to February 2013 (54 infants), intervention year 1 from March 2013 to February 2014 (61 infants), and intervention year 2 from March 2014 to February 2015 (48 infants)
	Outcomes	Decrease in use of pharmacological treatment (46% to 27%), adjunctive use of phenobarbital (13% to 2%), mean length of hospitalization for pharmacologically treated infants (16.9 days to 12.3 days), mean cost per treated infant (USD 19,737 to USD 8755; P < 0.01), mean cost per at risk patient (USD 11,000 to USD 5300; P < 0.01), and cumulative morphine exposure (13.7 mg to 6.6 mg per treated newborn) No change noted in length of hospitalization for non‐pharmacologically treated infants (4.2 days to 4.4 days; P = 0.33) No change in mean daily Finnegan score; quote: "no significant difference in median score, maximum score, or first score by year (P = 0.53, 0.29, 0.48, respectively)" No change in number of newborns discharged in parental custody (93% to 90%; P = 0.73)
	Notes	—
Kirchner 2014	Methods	Retrospective cohort
	Participants	131 infants who completed an international multicenter clinical trial ("MOTHER study")
	Time period	2005–2008
	Site of care	Unit not specified (international)
	Interventions	Comparison by site of enrollment; quote: "the 7 contributing sites were divided into 3 clusters: the European cluster (Vienna site, n=37), the rural US cluster (N=39): Burlington (University of Vermont, n=26), Nashville (Vanderbilt University, n=13), and the urban US cluster (n=55): Detroit (Wayne State University, n=12), Providence (Brown University, n=3), Philadelphia (Thomas Jefferson University, n=23), Baltimore (Johns Hopkins University, n=17)" Quote: "A standardized NAS rating and treatment protocol was applied, while non‐pharmacologic care of NAS symptoms differed across sites" Quote: "In Europe, all neonates (also those without treatment requirement) stayed in the hospital accommodated in a room with their mothers for at least 10 days" Quote: "The breastfeeding rate was much higher in Europe and was highest in rural US"
	Outcomes	Quote: "The lowest morphine doses for both medication conditions were administered in Europe where all babies were kept in a rooming‐in situation in hospital for at least 10 days" Pharmacological treatment rates: 18.5% Europe rooming in model vs 29.5% urban USA vs 18.5% rural USA Mean total dose of morphine within the first 28 postpartum days: 5.38 mg Europe rooming in model vs 35.05 mg urban USA vs 8.66 mg rural US Quote: "Urban US neonates had significantly higher NAS scores (p<0.01) compared to rural US and European neonates, and needed significantly higher morphine doses (P < 0.05) with longer treatment duration" Quote: "Urban US neonates exhibited most neurological symptoms (P < 0.001) while in Europe autonomous, respiratory and gastrointestinal symptoms were found significantly more often compared to urban and/or rural US (P < 0.05)"
	Notes	Quote: "No differences between sites could be found in behavioural symptoms, but interestingly a significant positive influence on these symptoms could be demonstrated by breastfeeding (β = –0.261; P = 0.034), with explaining 6.8% of variance"
Loudin 2017	Methods	Retrospective cohort (historical controls)
	Participants	1023; quote: "inborn neonates >35 weeks' gestational age with the diagnosis of NAS (ICD9‐CM 779.5), requiring pharmacologic treatment"
	Time period	2010–2015
	Site of care	Neonatal abstinence center, NTU, and NICU (West Virginia, USA)
	Interventions	Baseline (number of infants not reported); quote: "managed in the NICU, a 36 bed level III unit with an average of 550 admissions per year" and post‐implementation (number of infants not reported) of a neonatal therapeutic unit (NTU), (quote) "a 15‐bed low‐light, low‐noise unit with a dedicated nursing staff and nursing aids along with community volunteers called 'rockers' who were specially trained to hold and rock the neonates when appropriate and when family members were unavailable. The staff utilized therapeutic handling techniques that further enhanced calming effects on the neonates." Also compared to post‐implementation (78 infants) of an offsite; quote: "neonatal abstinence center with 12 beds and a dedicated staff of nurses, patient care assistants, and volunteers trained in the care of neonates with NAS. In addition, on‐site social service personnel served to transition caretakers (birth parents, foster parents or adopting families) … to a safe home environment. Limited number of patient rooms were available to accommodate rooming in, thus that space was utilized to help prepare caretakers prior to discharge."
	Outcomes	Quote: "The median [length of stay] for NAS medication‐treated patients in the NICU without comorbidities was 24 days (interquartile range 24 to 52). Median [length of stay] in the NTU was 26 days (interquartile range 26 to 42). The median [length of stay] for Lily's Place was 33 days (interquartile range 32 to 60 days), indicative of the transition to home life mission of the facility in addition to extended neonatal withdrawal care" Significantly higher median cost per patient for those managed in the NICU (USD 90,601, IQR USD 64,489 to USD 128,135) compared to the NTU (USD 68,750, IQR USD 44,952 to USD 92,548) and then compared to the offsite neonatal abstinence center (USD 17,688, IQR USD 9933 to USD 20,033); P < 0.0001 Quote: "The percentage of total NAS patients requiring pharmacologic treatment in the NICU fell from 100% prior to the establishment of the NTU to 22% by 2015"
	Notes	—
Miles 2007	Methods	Retrospective cohort (historical controls)
	Participants	176 infants of women in a methadone maintenance program
	Time period	1991–1994, 1997–2001
	Site of care	Postnatal ward and neonatal medicine unit (UK)
	Interventions	Baseline (78 infants) and postimplementation (98 infants) of a shared care approach in clinical management with a drug liaison midwife service for mothers and infants. Also changed from primary admission to the neonatal medicine unit to primary admission to the maternity ward with transfer to the neonatal medicine unit if pharmacological treatment was required. Additionally, scoring was modified, the threshold for pharmacological treatment adjusted, and treatment choice changed from chlorpromazine to phenobarbitone. Quote: "The [drug liaison midwife] provided specialist advice regarding methadone treatment, care of the newborn, and the advantages of breastfeeding."
	Outcomes	Quote: "Infants spend less time in hospital (median 5 days vs 28 days, P < 0.0002), a smaller proportion had treatment for NAS (14% vs 79%, P = 0.001), and neonatal medicine unit admission was reduced (median 14 days vs 26 days, P < 0.0003 [number of infants 100% vs 40%, P < 0.001])" Quote: "Neonatal convulsions (P = 0.0001) and jaundice (P < 0.001) occurred less frequently, and more infants were breastfed (P = 0.001)" Quote: "About 10% of infants were discharged to a foster family" in both time periods Quote: "None of these children had impairments of vision, hearing, growth, or development" in the first year after leaving the hospital (baseline n = 24, post‐implementation n = 98)
	Notes	—
Patrick 2016	Methods	Retrospective and prospective cohort (quality improvement methodology)
	Participants	3458 infants pharmacologically treated for NAS. Data were collected through serial cross‐sectional audits of 199 participating centers
	Time period	2012–2014
	Site of care	Unit not specified (International)
	Interventions	Quote: "Participating NICUs shared common interests, worked collaboratively, shared information and ideas, and developed standardized practices." The goal was to facilitate guideline update "for the following: (1) screening for maternal substance use; (2) nonpharmacologic treatment of infants with NAS; (3) scoring signs of NAS; (4) breastfeeding; (5) pharmacologic management; and (6) duration of observation of exposed infants."
	Outcomes	Quote: "(T)he mean number of NAS‐focused guidelines increased from 3.7 to 5.1 of a possible 6 (P < .001), with improvements noted in all measured domains" Quote: "(D)ecreases occurred in median (interquartile range) length of pharmacologic treatment, from 16 days (10 to 27 days) to 15 days (10 to 24 days; P = .02) and [length of stay] from 21 days (14 to 33 days) to 19 days (15 to 28 days; P = .002)" The difference in predicted mean length of treatment for hospitals that implemented policies about non‐pharmacological treatment of NAS was –2.0 days (95% CI –4.4 to 0.5) with a predicted length of stay of –2.5 days (95% CI –5.1 to 1.2) The difference in predicted mean length of treatment for hospitals that implemented policies about breastfeeding or provision of expressed human milk was –0.7 days (95% CI –2.1 to 0.7) with a predicted length of stay of –0.5 days (95% CI –2.1 to 1.1)
	Notes	—
Wachman 2018	Methods	Retrospective and prospective cohort (pre‐/postquality improvement methodology)
	Participants	240 opioid‐exposed infants > 36 weeks' gestation
	Time period	April 2015 to December 2017
	Site of care	Pediatrics unit and postnatal ward (Massachusetts, USA)
	Interventions	Baseline (101 infants), intervention (54 infants), and postintervention (85 infants). Interventions included (quote): "a non‐pharmacologic care bundle, function‐based assessments consisting of symptom prioritization and then the "Eat, Sleep, Console" (ESC) Tool; and a switch to methadone for pharmacologic treatment."
	Outcomes	Quote: "Pharmacologic treatment decreased from 87.1 to 40.0%; adjunctive agent use from 33.6 to 2.4%; hospitalization length from a mean 17.4 to 11.3 days, and opioid treatment days from 16.2 to 12.7 (P < 0.001 for all). Total hospital charges decreased from $31,825 to $20,668 per infant. Parental presence increased from 55.6 to 75.8% (P < 0.0001). No adverse events were noted." Hospital length of stay for pharmacologically treated infants trended downward (19.1 days to 17.6 days; P = 0.11) Quote: "There were no NICU admissions for management of NAS or its complications, and no seizures" No significant change in NICU admission (23.7% to 21.2%; P = 0.78) or DCF custody status (19.8% to 28.2%; P = 0.18)
	Notes	Quote: "Breastfeeding initiation rate did not differ during the two time periods."

CI: confidence interval; DCF: Department of Children and Families; ESC: Eat, Sleep, Console Tool; IQR: interquartile range; MOTHER: Maternal Opioid Treatment: Human Experimental Research; NAS: neonatal abstinence syndrome; NICU: neonatal intensive care unit; NTU: neonatal therapeutic unit.

Ongoing studies

We identified seven ongoing studies in the search that may qualify for inclusion at review update when complete (Characteristics of ongoing studies table; NCT02768844; NCT02801331; NCT03097484; NCT03113656; NCT03533985; NCT03549936; NCT03987165).

See Figure 1 for the selection process.

Results of the search

Comparison 1

We identified six RCTs comparing single or bundled non‐pharmacological interventions to no non‐pharmacological treatment or different single or bundled non‐pharmacological interventions in infants at risk for, or having early symptoms consistent with, opioid withdrawal as defined above (see Characteristics of included studies table).

Comparison 2

We identified no RCTs comparing single or bundled non‐pharmacological interventions to no non‐pharmacological treatment or different single or bundled non‐pharmacological interventions in infants receiving opioid treatment for symptoms consistent with opioid withdrawal as defined above.

Included studies

Modifying environmental stimulation

Ostrea 1975 compared 198 infants randomized to either a study nursery or the regular nursery at a children's hospital in the USA. This study included infants at risk for opioid withdrawal due to maternal opioid exposure (131 clinic infants, 57 non‐clinic infants). Non‐opioid exposure and feeding type were not specified. Infants in the study nursery were placed in incubators; fed every three hours; and maintained in a secluded, quiet, dim room. In comparison, infants in the regular nursery were placed in open bassinets, fed every four hours, and exposed to normal nursery noise and light. As measured by a center‐specific withdrawal rating, there were no significant differences in severity or pattern of withdrawal symptoms or in degree of weight loss. Eleven infants received pharmacological treatment and there was no difference between groups.

Oro 1988 compared 49 infants randomized to either non‐oscillating waterbeds or conventional bassinets in a nursery at a hospital in the USA. This study included infants at risk for opioid withdrawal due to prenatal narcotic exposure and heroin or methadone, or both, on urine toxicology results. Twelve narcotic‐exposed and three with coexposure to cocaine or methamphetamine, or both, were chosen from the intervention group and matched to comparable control neonates. Outcomes were not reported for 19 randomized infants. All infants were formula fed. Infants in the intervention group showed less severe central nervous system withdrawal symptoms, improved weight gain, required a lower maximum dose of phenobarbital, and more infants were weaned off from medication prior to discharge. There was no significant difference in overall use of pharmacological treatment.

D'Apolito 1999 compared 14 infants randomized to either a rocking bed or a standard bed from 24 hours of life to seven days of life in a nursery at a university hospital in the USA. This study included infants at risk for opioid withdrawal due to prenatal methadone exposure. Non‐opioid exposures included ethanol, heroin, marijuana, and cigarettes, but excluded cocaine. Feeding type was not reported. The rocking bed intervention consisted of a motor‐driven rocking bed set at 75% rocking that started and stopped at randomly determined intervals and was used in combination with maternal intrauterine sounds. Infants in the intervention group showed increased withdrawal symptoms by the seventh day of life and increased sleep disruption. All infants received pharmacotherapy for withdrawal symptoms.

Maichuk 1999 compared 48 infants randomized to either prone sleep positioning or supine sleep positioning in an intensive care nursery at a university hospital in the USA. This study included infants at risk for opioid withdrawal with urine toxicology positive for heroin or methadone, or both. Approximately 60% of infants in each group also had urine toxicology positive for cocaine. All infants were formula fed. All infants were swaddled for sleep. Infants in the intervention group were then placed prone. Infants in the intervention group had lower peak and mean withdrawal scores as well as lower caloric intake. All infants received pharmacotherapy for withdrawal symptoms.

Feeding practices

Bogen 2018 compared 49 infants randomized to either 24 kcal/oz formula or 20 kcal/oz formula from day three to 21 of life at a large, urban hospital in the USA. Site of care was not specified. This study included infants at risk for opioid withdrawal due to maternal methadone treatment. Maternal and infant drug screens at delivery were positive for illicit drugs more often in the control group. Both breastfed and formula‐fed infants were included in the study and breast milk feedings were prioritized over study formula feedings. The main outcome was evaluation of weight parameters. Infants in the high‐calorie intervention group trended toward improved weight parameters as measured by weight nadir, days to return to birth weight, and mean daily weight gain. A similar proportion of infants in each group received pharmacological treatment and there was no significant difference in treatment duration for those who used pharmacological treatment. There was significant crossover between groups.

Support of the mother‐infant dyad

MacVicar 2018 compared 14 mother‐infant dyads randomized to either tailored breastfeeding support or standard Baby‐Friendly Initiative care in a tertiary maternity hospital in Scotland. Infants were admitted to a room with their mothers and transferred to the neonatal unit for pharmacological treatment. This study included infants at risk for opioid withdrawal due to maternal opioid substitution therapy during pregnancy. Mothers with ongoing illicit psychoactive drug or alcohol use were excluded. All mothers intended to breastfeed. Tailored support included a one‐hour session daily for five days with a dedicated support worker. Sessions focused on breastfeeding support, promotion of self‐efficacy, neonatal self‐consolation techniques, and low‐stimuli environment. Infants in the intervention group were less likely to receive pharmacological treatment, and had a shorter length of stay compared to the control group. This effect may have been related to feeding practice since all infants in the intervention group were breastfed on day five compared to four of seven in the control group or to the ability to continue to room in with their mothers. In post hoc analysis of feeding practice, breastfed infants were less likely to be treated with pharmacological therapy and had a shorter length of stay.

Excluded studies

We excluded 34 non‐randomized studies in the search. These studies are described in a narrative presentation here. Further details about studies are included in and supplementary tables and the appendix (see Characteristics of excluded studies table; Table 1; Table 2; Table 3; Table 4; and Appendix 3).

We described cohort studies, case‐controlled studies, and cross‐sectional studies with a clear comparison of non‐pharmacological care practices. We did not describe studies without a clear comparison of non‐pharmacological care practices such as case series or case reports. We classified studies as prospective, retrospective, or both prospective and retrospective. We described studies using quality improvement methodology as prospective and retrospective cohorts. We classified studies as either historically or concurrently controlled.

Modifying environmental stimulation

(Adjustments of environmental stimulation include low light, low noise, clustered care, containment or swaddling, bedding type, positioning, non‐nutritive sucking, bathing, aromatherapy, music therapy, and massage.)

Two prospective feasibility studies explored the effect of modifying environmental stimulation on individual opioid‐exposed newborns. Radziewicz 2018 evaluated the effect of a 30‐minute Reiki treatment in a newborn nursery. Zuzarte 2017 evaluated the effect of stochastic vibrotactile stimulation delivered by a specially constructed mattress in 30‐minute intervals in the NICU and a newborn nursery. Both studies were conducted in the USA. Neither study recorded outcomes prespecified for this review.

Feeding practices

(Feeding modifications include infant‐led feeding, high caloric feedings, tube feedings, maternal breast milk feedings, non‐maternal breast milk feedings, low‐lactose formula, and breastfeeding.)

Fifteen studies assessed the effect of feeding practices on outcomes in opioid‐exposed newborns. Thirteen studies were cohort studies including 11 retrospective (Abdel‐Latif 2006; Arlettaz 2005; Dryden 2009; Hodgson 2012; Isemann 2011; Liu 2015; McQueen 2011; Metz 2015; O'Connor 2013; Pritham 2012; Short 2016), one retrospective and prospective (Welle‐Strand 2013), and one prospective study (Wachman 2013). Two additional studies used case‐control methodology, one with concurrent controls (Jansson 2008), and one with historical controls (Radmacher 2017). Twelve studies assessed infants at risk for, or having, early symptoms consistent with opioid withdrawal (Abdel‐Latif 2006; Arlettaz 2005; Dryden 2009; Hodgson 2012; Jansson 2008; Liu 2015; McQueen 2011; Metz 2015; O'Connor 2013; Pritham 2012; Wachman 2013; Welle‐Strand 2013); two studies assessed infants receiving opioid treatment for symptoms consistent with opioid withdrawal (Isemann 2011; Radmacher 2017); and one study did not collect information about postnatal treatment (Short 2016). Seven studies were conducted in the USA (Isemann 2011; Jansson 2008; O'Connor 2013; Pritham 2012; Radmacher 2017; Short 2016; Wachman 2013); two in Australia (Abdel‐Latif 2006; Liu 2015); two in Canada (Hodgson 2012; McQueen 2011); and one each in the UK (Dryden 2009), Switzerland (Arlettaz 2005), Austria (Metz 2015), and Norway (Welle‐Strand 2013). One study evaluated infants cared for in the postnatal ward for monitoring and pharmacological treatment (Hodgson 2012), two studies evaluated infants cared for in the postnatal ward with transfer to the NICU for pharmacological treatment (Abdel‐Latif 2006; Dryden 2009), three studies evaluated infants cared for in the NICU (Isemann 2011; Pritham 2012; Radmacher 2017), and nine studies did not report the site of care (Arlettaz 2005; Jansson 2008; Liu 2015; McQueen 2011; Metz 2015; O'Connor 2013; Short 2016; Wachman 2013; Welle‐Strand 2013). All studies assessed the impact of breast milk feedings compared to non‐breast milk feedings on outcomes. Intervention definitions varied between studies in terms of timing and 'dose' of breast milk. Only one study assessed the impact of non‐maternal donor breast milk feedings (Radmacher 2017). Measured primary outcomes included length of hospitalization (Abdel‐Latif 2006; Dryden 2009; Isemann 2011; Metz 2015; O'Connor 2013; Pritham 2012; Radmacher 2017; Short 2016; Wachman 2013), use of pharmacological treatment (Abdel‐Latif 2006; Dryden 2009; Hodgson 2012; Jansson 2008; Liu 2015; O'Connor 2013; Wachman 2013; Welle‐Strand 2013), length of pharmacological treatment (Abdel‐Latif 2006; Jansson 2008; Metz 2015; Radmacher 2017; Welle‐Strand 2013), maximum dose of opioid medication (Abdel‐Latif 2006; Liu 2015), and cumulative dose of opioid medication (Metz 2015). One study evaluated the secondary outcomes of growth rate and days to regain birth weight (Radmacher 2017), another evaluated weight nadir (Dryden 2009), and one evaluated custody at discharge (Abdel‐Latif 2006). Though no studies assessed the predefined secondary outcome of peak NAS score in the first 72 hours of life, eight studies assessed the severity of NAS using other measures (Abdel‐Latif 2006; Arlettaz 2005; Jansson 2008; Liu 2015; McQueen 2011; Metz 2015; O'Connor 2013; Radmacher 2017).

Support of the mother‐infant dyad

(Treatments focused on the mother‐infant dyad include parental presence, rooming in, skin‐to‐skin or 'kangaroo care,' parental respite or 'cuddler' programs, trauma‐informed care, social work support, mental health support, treatment program support, and legal support.)

Ten studies evaluated methods of support for the mother‐infant dyad on outcomes in opioid‐exposed newborns. Nine were retrospective cohort studies, seven with concurrent controls (Abrahams 2007; Abrahams 2010; Howard 2017; Hünseler 2013; Metz 2011; Ordean 2015; Summey 2018), and two with historical controls (McKnight 2016; Saiki 2010). One study was a retrospective and prospective cohort utilizing quality improvement methodology (Crook 2017). Nine studies assessed infants at risk for, or having, early symptoms consistent with opioid withdrawal (Abrahams 2007; Abrahams 2010; Crook 2017; Hünseler 2013; McKnight 2016; Metz 2011; Ordean 2015; Saiki 2010; Summey 2018), and one study assessed infants receiving opioid treatment for symptoms consistent with opioid withdrawal (Howard 2017). Four studies were conducted in Canada (Abrahams 2007; Abrahams 2010; McKnight 2016; Ordean 2015); three in the USA (Howard 2017; Ordean 2015; Summey 2018); and one each in the UK (Saiki 2010), Austria (Metz 2011), and Germany (Hünseler 2013). Four studies evaluated infants cared for in the postnatal ward compared to a NICU setting (Abrahams 2007; Abrahams 2010; Saiki 2010; Summey 2018), two studies evaluated infants cared for in a pediatrics ward (Howard 2017; McKnight 2016), two studies evaluated infants cared for in the NICU (Crook 2017; Hünseler 2013), and two studies did not report the site of care (Metz 2011; Ordean 2015). Eight studies assessed the impact of rooming in, five as the primary intervention (Abrahams 2007; Hünseler 2013; McKnight 2016; Ordean 2015; Saiki 2010), and three in combination with additional support and education (Abrahams 2010; Metz 2011; Summey 2018). One study examined a breastfeeding education program (Crook 2017), and another examined the impact of parental presence (Howard 2017). Measured primary outcomes included length of hospitalization (Abrahams 2007; Abrahams 2010; Crook 2017; Howard 2017; Hünseler 2013; McKnight 2016; Metz 2011; Ordean 2015; Saiki 2010; Summey 2018), use of pharmacological treatment (Abrahams 2007; Crook 2017, Hünseler 2013, McKnight 2016; Metz 2011; Ordean 2015; Saiki 2010), length of pharmacological treatment (Abrahams 2007; Howard 2017; Hünseler 2013; McKnight 2016; Metz 2011; Ordean 2015; Saiki 2010), maximum dose of opioid medication (Hünseler 2013; McKnight 2016), and cumulative dose of opioid medication (Howard 2017; Metz 2011). Secondary outcomes assessed included NICU admission (Abrahams 2007; Abrahams 2010; Ordean 2015; Summey 2018), length of NICU stay (Abrahams 2010), weight nadir (Abrahams 2007; Ordean 2015; Summey 2018), seizure (Summey 2018), custody at discharge (Abrahams 2007; Abrahams 2010; Howard 2017; Hünseler 2013; Ordean 2015; Saiki 2010; Summey 2018), and cost of care (Hünseler 2013; Summey 2018). Though no studies assessed the predefined secondary outcome of peak NAS score in first 72 hours of life, two studies assessed the severity of NAS using other measures (Howard 2017; Hünseler 2013).

Multiple non‐pharmacological care practices

(Any combination of the above non‐pharmacological interventions.)

Seven studies examined multiple non‐pharmacological care practices in combination with other non‐pharmacological and pharmacological treatments. Four studies were retrospective and prospective cohorts utilizing quality improvement methodology (Grossman 2017; Holmes 2016; Patrick 2016; Wachman 2018). Three studies were retrospective cohort studies, one with concurrent controls (Kirchner 2014), and two with historical controls (Loudin 2017; Miles 2007). Five studies assessed infants at risk for, or having, early symptoms consistent with opioid withdrawal (Grossman 2017; Holmes 2016; Kirchner 2014; Miles 2007; Wachman 2018), and two studies assessed infants receiving opioid treatment for symptoms consistent with opioid withdrawal (Loudin 2017; Patrick 2016). Four studies were conducted in the USA (Grossman 2017; Holmes 2016; Loudin 2017; Wachman 2018), two at multiple international sites (Kirchner 2014; Patrick 2016), and one in the UK (Miles 2007). Two studies evaluated infants cared for in the postnatal ward, pediatrics unit, and NICU settings (Grossman 2017; Holmes 2016); one study evaluated infants cared for in the postnatal ward and the pediatrics unit (Wachman 2018); one study evaluated infants cared for in the postnatal ward and the NICU (Miles 2007); and two studies did not report the site of care (Kirchner 2014; Patrick 2016). Three studies assessed a combination of standardized non‐pharmacological care, education, and rooming in among other interventions (Grossman 2017; Holmes 2016; Wachman 2018). Two studies compared different care environments (Kirchner 2014; Loudin 2017). One study examined the combination of rooming in, maternal support, and changes in pharmacological care practices in combination (Miles 2007). One study evaluated the impact of a multicenter effort to standardize care and guidelines affecting both non‐pharmacological and pharmacological care practices (Patrick 2016). Measured primary outcomes included length of hospitalization (Grossman 2017; Holmes 2016; Loudin 2017; Miles 2007; Patrick 2016; Wachman 2018), use of pharmacological treatment (Grossman 2017; Holmes 2016; Kirchner 2014; Miles 2007; Wachman 2018), length of pharmacological treatment (Kirchner 2014; Patrick 2016; Wachman 2018), and cumulative dose of opioid medication (Holmes 2016; Kirchner 2014). Secondary outcomes evaluated include NICU admission (Grossman 2017; Loudin 2017; Miles 2007; Wachman 2018), seizures (Miles 2007; Wachman 2018), custody at discharge (Holmes 2016; Miles 2007), and cost of care (Grossman 2017; Holmes 2016; Loudin 2017). Though no studies assessed the predefined secondary outcome of peak NAS score in first 72 hours of life, two studies assessed the severity of NAS using other measures (Holmes 2016; Kirchner 2014). No studies assessed neurodevelopmental outcome as prespecified in the protocol at two years; however, one study assessed neurodevelopmental outcome in the first year of life (Miles 2007).

Risk of bias in included studies

Six RCTs are included in a risk of bias assessment (Figure 2).

Figure 2

Risk of bias summary: review authors' judgments about each risk of bias item for each included study.

Modifying environmental stimulation (D'Apolito 1999; Maichuk 1999; Oro 1988; Ostrea 1975).
Feeding practices (Bogen 2018).
Support of the mother‐infant dyad (MacVicar 2018).

The six studies demonstrated a generally high risk of bias. Bogen 2018 exhibited low risk of bias in several categories; however, sensitivity analysis was not performed because it was the only study in the category.

Allocation

Modifying environmental stimulation: we judged Oro 1988 to have high risk of sequence generation bias due to randomization of infants based on hospital record numbers. Three studies did not describe the randomization process and we judged risk of bias due to sequence generation as unclear (D'Apolito 1999; Maichuk 1999; Ostrea 1975). None of the four studies report blinded allocation and we judged risk of selection bias as unclear.

Feeding practices: we judged Bogen 2018 to have low risk of sequence generation and selection bias due to randomization of infants with blinded computer‐based sequence generation.

Support of the mother‐infant dyad: we judged MacVicar 2018 to have low risk sequence generation bias due to randomization of infants with computer‐based sequence generation. Blinded allocation was not reported, and we judged risk of selection bias as unclear.

Blinding

Modifying environmental stimulation: four studies did not blind caregivers or evaluators to the intervention due to inability to conceal the interventions during routine care. We judged the risk of performance and detection bias as high (D'Apolito 1999; Maichuk 1999; Oro 1988; Ostrea 1975).

Feeding practices:Bogen 2018 blinded all caregivers and evaluators to the intervention. We judged the risk of performance and detection bias as low.

Support of the mother‐infant dyad:MacVicar 2018 did not blind caregivers or evaluators to the intervention due to inability to conceal the interventions during routine care. We judged the risk of performance and detection bias as high.

Incomplete outcome data

Modifying environmental stimulation: we judged Oro 1988 to have a high risk of attrition bias due to the exclusion of randomized infants who were unable to be matched to controls (approximately 39% of randomized infants). Three studies reported outcomes for all enrolled infants, and we judged the risk of attrition bias as low (D'Apolito 1999; Maichuk 1999; Ostrea 1975).

Feeding practices: we judged Bogen 2018 to have a high risk of attrition bias due to the proportion of infants lost to follow‐up before reaching birth weight or the end of the study intervention on day 21 (18%).

Support of the mother‐infant dyad: we judged MacVicar 2018 to have a high risk of attrition bias due to the proportion of infants stopping the intervention early (approximately 43%), and the loss to follow‐up posthospitalization (approximately 21%).

Selective reporting

Modifying environmental stimulation: all four studies were not preregistered (study predated the mandatory requirement for preregistration) and there was no clear evidence of selective reporting (D'Apolito 1999; Maichuk 1999; Oro 1988; Ostrea 1975). We judged the risk of reporting bias as unclear.

Feeding practices:Bogen 2018 was preregistered; however, several prespecified outcomes were not reported in the publication raising the possibility of selective reporting. We judged the risk of reporting bias as high.

Support of the mother‐infant dyad:MacVicar 2018 was not preregistered (expected for the publication year) and there was no clear evidence of selective reporting. We judged the risk of reporting bias as unclear.

Other potential sources of bias

Modifying environmental stimulation:D'Apolito 1999 and Oro 1988 noted differences in the number of infants with fetal distress. Maichuk 1999 and Ostrea 1975 did not report differences between treatment groups. None of the studies reported the number of eligible infants not enrolled and we could not assess the risk for differential enrollment affecting generalizability of study results. We judged the risk of other potential sources of bias in these studies as unclear.

Feeding practices:Bogen 2018 noted crossover and protocol deviation between groups that could have biased the results toward the null. Six infants assigned to standard‐calorie formula and one assigned to high‐calorie formula changed to open‐label high‐calorie formula due to slow weight gain. The change was made per protocol for two infants, and determined by non‐study physicians for four infants. Three infants assigned to high‐calorie formula switched to soy formula based on physician or family preference. Primary analysis was performed based on original intention‐to‐treat analysis. Post hoc analysis was performed based on feeding type. Bogen 2018 also noted baseline differences between the intervention and control groups. Six infants had positive screens at delivery for drugs other than methadone in the intervention group with no exposure in the control group. Bogen 2018 did not enroll 35% of infants whose mothers enrolled during pregnancy, largely due to need for prolonged NICU care not related to NAS and low birth weight. This may affect generalizability of study results. We judged the risk of other potential sources of bias as high.

Support of the mother‐infant dyad:MacVicar 2018 noted baseline differences between the intervention and control groups. There were more multiparous women and women with experience breastfeeding in the intervention group compared to the control group, and this difference could have a significant effect on a mother's initial comfort with newborn care. MacVicar 2018 did not enroll 74% of mothers approached, suggesting the study group may be a non‐representative population and likely a self‐selected population with intent to breastfeed. This may affect generalizability of study results. We judged the risk of other potential sources of bias as high.

Effects of interventions

Comparison 1. Infants at risk for, or having early symptoms consistent with, opioid withdrawal

See summary of findings Table 1.

Modifying environmental stimulation

Four studies reported modifying environmental stimulation (D'Apolito 1999; Maichuk 1999; Oro 1988; Ostrea 1975).

The studies compared a study nursery to a regular nursery (Ostrea 1975), non‐oscillating waterbeds to conventional bassinets (Oro 1988), rocking beds to standard beds (D'Apolito 1999), and prone sleep positioning to supine sleep positioning (Maichuk 1999).

Primary outcomes

Length of hospitalization (days) (outcome 1.1.1)

Oro 1988 reported a mean reduction in length of hospitalization by 1 day with a wide 95% CI ranging from a reduction of 2.82 days to an increase of 0.82 days (30 infants) (Analysis 1.1). We judged the certainty of evidence as very low and downgraded three levels for very serious concerns about risk of bias (due to unblinded study, concern for performance, detection, and attrition bias) and serious concern for imprecision (due to wide 95% CI).

Pharmacological treatment with one or more doses of opioid or sedative medication (outcome 1.2.1)

Three studies of modifying environmental stimulation reported no difference in the use of pharmacological treatment with one or more doses of opioid or sedative medication (D'Apolito 1999; Maichuk 1999; Oro 1988). Meta‐analysis of the three studies found no evidence of a difference in the use of pharmacological treatment (typical RR 1.00, 95% CI 0.86 to 1.16; I² = 0; 92 infants; Analysis 1.2). We judged the certainty of evidence as low and downgraded two levels for very serious concerns about risk of bias (due to unblinded study, concern for performance, detection, and attrition bias).

Secondary outcomes

Peak neonatal abstinence syndrome score in first 72 hours of life (outcome 1.3.1)

None of the studies of modifying environmental stimulation reported the peak NAS score in the first 72 hours of life as prespecified in the protocol. Post hoc we decided to evaluate 'peak' score at any time or any score reported in the first 72 hours as measured by a standardized withdrawal assessment tool (Analysis 1.3).

Maichuk 1999 studied infant positioning during sleep and reported a mean reduction of peak neonatal abstinence score at any time of 2.65 (95% CI –3.81 to –1.49; 48 infants). D'Apolito 1999 and Oro 1988 studied bed type. D'Apolito 1999 reported the mean score in the first 72 hours of life (MD 1.60, 95% CI –1.11 to 4.31; 14 infants) and Oro 1988 reported peak neonatal abstinence score at any time (MD –0.20, 95% CI –0.82 to 0.42; 30 infants). Neither demonstrated a reduction of score. Meta‐analysis of the three studies of peak NAS score demonstrated extremely high heterogeneity and are not reported.

Neonatal intensive care unit admission (outcome 1.4.1)

None of the studies reported the effect of modifying environmental stimulation on NICU admission.

Days to regain birth weight (outcome 1.5.1)

None of the studies reported the effect of modifying environmental stimulation on days to regain birth weight.

Weight nadir (outcome 1.6.1)

Ostrea 1975 reported a mean reduction in weight nadir by 0.28% with a wide 95% CI ranging from a reduction of 1.15% to an increase of 0.59% (194 infants) (Analysis 1.6). We judged the certainty of evidence as very low and downgraded three levels for very serious concerns about risk of bias (due to unblinded study, concern for performance and detection bias) and serious concern for imprecision (due to wide 95% CI).

Feeding practices

One study reported feeding practices (Bogen 2018).

The study compared 24 kcal/oz formula to 20 kcal/oz formula.

Primary outcomes

Length of hospitalization (days) (outcome 1.1.2)

The study did not report the effect of feeding practices on length of hospitalization.

Pharmacological treatment with one or more doses of opioid or sedative medication (outcome 1.2.2)

Bogen 2018 reported a reduced use of pharmacological treatment with an RR of 0.92 and a wide 95% CI ranging from 0.63 to 1.33 (49 infants) (Analysis 1.2). We judged the certainty of evidence as very low and downgraded three levels for very serious concerns about risk of bias (due to concern for attrition and reporting bias), serious concern about imprecision (due to wide 95% CI), and plausible confounding that would reduce demonstrated effect.

Secondary outcomes

Peak neonatal abstinence syndrome (NAS) score in first 72 hours of life (outcome 1.3.2)

The study did not report the effect of feeding practices on peak NAS.

Neonatal intensive care unit admission (outcome 1.4.2)

The study did not report the effect of feeding practices on NICU admission.

Days to regain birth weight (outcome 1.5.2)

Bogen 2018 reported a reduction in the mean number of days to regain birth weight by 1.1 days with a wide 95% CI ranging from a reduction of 2.76 days to an increase of 0.56 days (46 infants) (Analysis 1.5). We judged the certainty of evidence as very low and downgraded three levels for very serious concerns about risk of bias (due to concern for attrition and reporting bias), very serious concern about imprecision (due to very wide 95% CI), and plausible confounding that would reduce demonstrated effect.

Weight nadir (outcome 1.6.2)

Bogen 2018 reported a mean reduction in weight nadir by 0.80% with a wide 95% CI ranging from a reduction of 2.24% to an increase of 0.64% (46 infants) (Analysis 1.6). We judged the certainty of evidence as very low and downgraded three levels for very serious concerns about risk of bias (due to concern for attrition and reporting bias), serious concern about imprecision (due to wide 95% CI), and plausible confounding that would reduce demonstrated effect.

Support of the mother‐infant dyad

One study reported support of the mother‐infant dyad (MacVicar 2018).

The study compared tailored breastfeeding support to standard Baby‐Friendly Initiative care.

Primary outcomes

Length of hospitalization (days) (outcome 1.1.3)

MacVicar 2018 reported a mean reduction in length of hospitalization by 8.9 days with a wide 95% CI ranging from a reduction of 19.84 days to an increase of 2.04 days (14 infants) (Analysis 1.1). We judged the certainty of evidence as very low and downgraded three levels for very serious concerns about risk of bias (due to unblinded study, concern for performance, detection, and attrition bias) and very serious concern for imprecision (due to very wide 95% CI).

Pharmacological treatment with one or more doses of opioid or sedative medication (outcome 1.2.3)

MacVicar 2018 reported a reduced use of pharmacological treatment with an RR of 0.50 and a wide 95% CI ranging from 0.13 to 1.90 (14 infants) (Analysis 1.2). We judged the certainty of evidence as very low and downgraded three levels for very serious concerns about risk of bias (due to unblinded study, concern for performance, detection, and attrition bias) and very serious concern for imprecision (due to very wide 95% CI).

Secondary outcomes

Peak neonatal abstinence syndrome (NAS) score in first 72 hours of life (outcome 1.3.3)

The study did not report the effect of support of the mother‐infant dyad on peak NAS.

Neonatal intensive care unit admission (outcome 1.4.3)

MacVicar 2018 reported a reduced need for NICU admission with an RR of 0.50 and a wide 95% CI ranging from 0.13 to 1.90 (14 infants) (Analysis 1.4). We judged the certainty of evidence as very low and downgraded three levels for very serious concerns about risk of bias (due to unblinded study, concern for performance, detection, and attrition bias) and very serious concern for imprecision (due to very wide 95% CI).

Days to regain birth weight (outcome 1.5.3)

The study did not report the effect of support of the mother‐infant dyad on days to regain birth weight.

Weight nadir (outcome 1.6.3)

The study did not report the effect of support of the mother‐infant dyad on weight nadir.

Other secondary outcomes for studies of infants at risk for, or having early symptoms consistent with, opioid withdrawal

None of the studies reported on length of NICU stay, neonatal seizure requiring anticonvulsant therapy, neonatal and infant all‐cause mortality, cerebral palsy, neurodevelopmental outcome at approximately two years' corrected age (cerebral palsy, developmental delay [Bayley or Griffith assessment more than two SD below mean] or intellectual impairment [IQ more than two SD below mean], blindness [vision less than 6/60 in both eyes], sensorineural deafness requiring amplification), complications of therapy, measures of maternal mood and bonding including Edinburgh Postpartum Depression Scale, Patient Health Questionnaire (2 or 9), Maternal Attachment Inventory, Maternal Postpartum Attachment Scale, Mother‐Infant Bonding Scale, or Postpartum Bonding Questionnaire (Brockington 2006; Condon 1998; Cox 1987; Kroenke 2001; Muller 1994; Taylor 2005; Whooley 1997), custody status at discharge, or cost of care.

Comparison 2. infants receiving opioid treatment for symptoms consistent with opioid withdrawal

We identified no studies.

Discussion

Summary of main results

We identified six RCTs evaluating 353 infants for inclusion in the analysis. The RCTs were published between 1975 and 2018 (six for comparison one, and none for comparison two). We identified and excluded 34 non‐randomized studies published between 2005 and 2018. We identified and excluded seven preregistered interventional clinical trials. These studies may qualify for inclusion at review update when complete. The certainty of evidence for the outcomes in the RCTs was very low to low. This was due to risk of bias inherent in the intervention, imprecision related to study size, and heterogeneity in study design.

Four studies evaluated modifying environmental stimulation by comparing a study nursery to a regular nursery (Ostrea 1975), non‐oscillating waterbeds to conventional bassinets (Oro 1988), rocking beds to standard beds (D'Apolito 1999), and prone sleep positioning to supine sleep positioning (Maichuk 1999). We are uncertain whether modifying environmental stimulation is associated with length of hospitalization based on one study with 30 infants (Oro 1988). Modifying environmental stimulation may be associated with little or no difference in use of pharmacological treatment based on three studies with 92 infants (D'Apolito 1999; Maichuk 1999; Ostrea 1975). We are uncertain whether modifying environmental stimulation is associated with peak NAS score at any time based on three studies with 92 infants (D'Apolito 1999; Maichuk 1999; Ostrea 1975). We are uncertain whether modifying environmental stimulation is associated with weight nadir based on one study with 194 infants (Ostrea 1975).

One study evaluated feeding practices by comparing 24 kcal/oz formula to 20 kcal/oz formula (Bogen 2018). We are uncertain whether feeding practices are associated with use of pharmacological treatment, days to regain birth weight, or weight nadir based on one study with 46 infants (Bogen 2018).

One study evaluated support of the mother‐infant dyad by comparing tailored breastfeeding support to standard Baby‐Friendly Initiative care (MacVicar 2018). We are uncertain whether support of the mother‐infant dyad is associated with length of hospitalization, use of pharmacological treatment, or NICU admission based on one study with 14 infants (MacVicar 2018).

Many prespecified outcomes were not reported. None of the studies reported length of NICU stay, neonatal seizures requiring anticonvulsant therapy, neonatal and infant all‐cause mortality, cerebral palsy, neurodevelopmental outcome at approximately two years' corrected age (cerebral palsy, developmental delay [Bayley or Griffith assessment more than two SD below mean] or intellectual impairment [IQ more than two SD below mean], blindness [vision less than 6/60 in both eyes], sensorineural deafness requiring amplification), complications of therapy, measures of maternal mood and bonding including Edinburgh Postpartum Depression Scale, Patient Health Questionnaire (2 or 9), Maternal Attachment Inventory, Maternal Postpartum Attachment Scale, Mother‐Infant Bonding Scale, or Postpartum Bonding Questionnaire (Brockington 2006; Condon 1998; Cox 1987; Kroenke 2001; Muller 1994; Taylor 2005; Whooley 1997), custody status at discharge, or cost of care. No studies evaluating modifying environmental stimulation looked at NICU admission or days to regain birth weight. No studies evaluating feeding practices looked at length of hospitalization, peak NAS score at any time, or NICU admission. No studies evaluating support of the mother‐infant dyad looked at peak NAS score at any time, days to regain birth weight, or weight nadir.

In summary, we identified few small studies of low quality and many prespecified outcomes were not reported. We are uncertain whether non‐pharmacological care for opioid withdrawal in newborns affects the prespecified outcomes. Given the low‐ or very low‐certainty evidence for all outcomes, the included studies do not support a specific non‐pharmacological intervention or combination of interventions over others.

Overall completeness and applicability of evidence

We identified six RCTs with heterogeneous study population, interventions, and outcome measures. The heterogeneity limits our ability to address the initial research question as well as generalizability of the results to other sites of care. Baseline outcome measures varied between studies suggesting that there are significant undefined differences between study sites. There is also lack of standardization in outcome measures and important measures are not measured in all studies (Kelly 2020).

The population studied is heterogeneous due to the presence of variable non‐opioid exposures. There is also high risk for differential enrollment not representative of the general population in this vulnerable population (Stine 2009). Study publications ranged from 1975 to 2018, and both substance exposures and treatments have changed over time. Many of the older studies were conducted when single drug exposure was common (e.g. heroin), and the model of care was more restrictive with fewer opportunities for parent contact, breastfeeding, or other aspects of non‐pharmacological care (Kocherlakota 2020). All but one of the RCTs was conducted in the USA, potentially limiting generalizability to other countries. Important demographic information such as, exact substance exposures, coexposures, and maternal medical and psychiatric history was not consistently collected.

We identified few studies in each category of non‐pharmacological intervention and heterogeneity of interventions within categories. In many of the included studies, there is lack of clarity in the 'dose', both in quality and quantity, of the intended intervention. The small number of studies and heterogeneity limited our ability to combine and compare studies directly. The identified studies lacked sufficient similarity for meta‐analysis for all outcomes except for the effect of modifying environmental stimulation on the use of pharmacotherapy. Though combined, these meta‐analyses include different interventions categorized as modifying environmental stimulation, including a waterbed, a rocking bed, and prone positioning.

The included studies showed differences in baseline outcome values between studies with no clear pattern. Due to lack of detail, we were unable to explore this source of heterogeneity between studies fully. The differences may be due to undefined cointerventions, which could include pharmacological treatment thresholds, non‐pharmacological interventions, and other multifactorial differences between study sites. Baseline and co‐occurring non‐pharmacological interventions were poorly defined in the studies. Clinically, cointerventions are common and some non‐pharmacological care practices are typically accompanied by other non‐pharmacological care practices. For example, breastfeeding is often accompanied by increases in clustered care, holding, skin‐to‐skin or 'kangaroo care', rocking, parental presence, and potentially additional support from staff (Bogen 2019). The practice of rooming in is often associated with increased rates of breastfeeding and may be associated with differences in environmental stimulation (Whalen 2019). These cointerventions, or bundles of non‐pharmacological interventions, are not always clearly defined or measured. Within the hospital, nursing ratios and site of care are likely significant unmeasured confounders that affect the ability to deliver non‐pharmacological interventions (Smith 2018). A specific site of care may be the most effective and efficient way to deliver certain non‐pharmacological interventions or combinations of non‐pharmacological interventions because site of care is highly interrelated with the ability to deliver non‐pharmacological care (Whalen 2019). For example, one intensive care unit may be unable to offer rooming in and have limited ability to decrease environmental stimulation. However, these limitations are not universal, and another intensive care unit may offer private rooms, intensive lactation and social support, and specific non‐pharmacological therapies. For this review, we chose to incorporate these differences between sites by assessing and comparing the specific interventions offered in each site studied. Limited data on cointerventions, site of care, and baseline confounding affect the generalizability of the study results.

While these studies assessed many predefined short‐term outcomes, they were not addressed consistently and important outcomes, such as NICU admission, were often missing. Heterogeneity of outcome measures in studies of neonates experiencing opioid withdrawal is common (Shan 2020). The largest and oldest study only contributed one prespecified secondary outcome of limited clinical importance. No studies addressed neonatal seizures requiring anticonvulsant therapy, all‐cause mortality, neurodevelopmental outcome, complications of therapy, measures of maternal mood and bonding, and custody at discharge.

Overall, based on six small RCTs, the completeness of evidence and applicability to the study question were limited by study heterogeneity, unmeasured baseline confounders and cointerventions, and unmeasured outcomes.

We also identified and excluded 34 non‐randomized studies published between 2005 and 2018 (29 for comparison 1, and five for comparison 2). The evidence is at baseline very low‐certainty based on the non‐randomized study design. The non‐randomized studies were identified using the standardized search strategies for randomized trials with additional studies identified when cross‐referencing for additional randomized studies. The search strategies and criteria for description of non‐randomized studies are less rigorous than those of RCTs. For this reason, we consider there is high likelihood that all eligible non‐randomized studies were not identified in the search process.

While the non‐randomized evidence is not methodologically strong, the 34 studies variably suggest improved outcomes with non‐pharmacological care practices when implemented singly or in combination. The majority of the non‐randomized studies showed either a decrease or no change in length of hospitalization, use of pharmacological treatment, and length of treatment when non‐pharmacological interventions were systematically implemented. The most widely studied non‐pharmacological care practices included breastfeeding, rooming in, and multiple non‐pharmacological practices implemented as part of a quality improvement initiative. Similar to the identified RCTs, the baseline outcome values and differences between groups varied significantly between sites for all interventions. This suggests that there are many unmeasured variables that impact care. The differences are likely partially related to unmeasured baseline differences in non‐pharmacological cointerventions and other multifactorial differences at each study site. There is also variability in quality and quantity of non‐pharmacological interventions. Heterogeneity in study design and size contribute to further variability.

Similar to the identified RCTs, the completeness of evidence and applicability to the study question based on non‐randomized studies are limited by study heterogeneity, unmeasured baseline confounders and cointerventions, and unmeasured outcomes.

Quality of the evidence

We judged the certainty of evidence for all outcomes in the six included RCTs as very low to low. Risk of bias was very serious for all studies secondary to the high risk of performance and detection bias related to lack of blinding (D'Apolito 1999; MacVicar 2018; Maichuk 1999; Oro 1988; Ostrea 1975), concern for attrition bias (Bogen 2018; MacVicar 2018; Oro 1988), concern for reporting bias (Bogen 2018), concern for other bias related to differential enrollment (Bogen 2018; MacVicar 2018), as well as many areas of unclear bias. Many outcomes were also downgraded for imprecision due to small study sizes with wide CIs. Though three studies reported the outcome of peak NAS score or NAS score in first 72 hours of life, meta‐analysis was not performed, and the outcome was downgraded due to serious inconsistency with marked heterogeneity and serious indirectness of the outcome measure. All outcomes for feeding measures were further downgraded for plausible confounding that would bias the results toward the null due to the significant number of participants who switched intervention groups in Bogen 2018.

Potential biases in the review process

Non‐pharmacological care for opioid withdrawal in newborns encompasses a group of interventions that are not well defined in the literature. The lack of standard terminology or definitions is a significant source of bias in the review process. For this review, we utilized a categorization of non‐pharmacological interventions based on common interventions reported in treatment of opioid withdrawal in newborns (Mangat 2019). There is no established standard research classification of non‐pharmacological care for opioid withdrawal in newborns and we established search terms based on our knowledge of the literature. Studies may categorize similar interventions in different ways such that the search terms did not identify all eligible studies. Additionally, some interventions such as breastfeeding or breast milk feedings are often examined as a subanalysis of a larger study that may not have been captured in the initial search. We included studies in categories of individual or combined non‐pharmacological interventions based on the study description and may have categorized studies incorrectly if the study description lacked detail. Within a category, we assumed the class and intensity of intervention was similar in the studies. When not described, we assumed a constant dose of each non‐pharmacological intervention.

The literature search using this categorization may not have captured all studies of the heterogeneous and historically not well‐defined group of non‐pharmacological interventions. Despite our best efforts, incomplete identification of studies for this review is possible. We reviewed 171 full‐text studies, many abstracts were advanced to full‐text review due to lack of clarity in study abstract about intervention and study design. We identified 28 of the full‐text studies outside of the search, largely in the references of other studies. The number of studies that were not captured by the search terms supports our suspicion for incomplete identification of studies. The majority of the full‐text studies reviewed were excluded due to study design, and it is likely that missed studies were a non‐randomized study design and would not qualify for inclusion. Importantly, the last search was performed in October 2019 and there may be more recent studies that qualify for inclusion.

Agreements and disagreements with other studies or reviews

While a number of reviews have assessed the impact of non‐pharmacological care practices together or as subgroups, we found no combined systematic review and meta‐analysis that evaluate non‐pharmacological care practices. Unlike many reviews of non‐pharmacological care practices, we chose to exclude studies of acupuncture and acupressure. A separate Cochrane Review is currently planned to address acupuncture and acupressure for opioid withdrawal in newborns, the protocol is not yet published. Similar to other recent reviews, we identified few RCTs and a larger literature of non‐randomized studies (MacMillan 2018; MacVicar 2019; Mangat 2019; McQueen 2019; Oostlander 2019; Ryan 2019). Unlike these reviews, we excluded non‐randomized studies. Our conclusion based on six RCTs was that we are uncertain whether non‐pharmacological care for opioid withdrawal in newborns affects the prespecified outcomes.

Other reviews that included non‐randomized studies discussed a trend toward improved outcomes when non‐pharmacological care practices were purposefully implemented. MacMillan 2018 noted reduced use of pharmacotherapy, length of stay, and cost of care in a systematic review and meta‐analysis of six non‐randomized studies evaluating rooming in. McQueen 2019 noted reduced incidence and duration of pharmacological treatment, shorter hospital length of stay, and decreased severity of NAS in methadone‐exposed infants in one randomized study (MacVicar 2018), and seven non‐randomized studies evaluating newborn feeding methods, the effect in buprenorphine‐exposed infants was unclear. MacVicar 2019 noted reduced use of pharmacotherapy in 10 studies and reduced length of stay in nine studies in a narrative review of non‐pharmacological care including 14 studies of heterogeneous non‐pharmacological interventions (13 non‐randomized, 1 RCT evaluating acupuncture; 2 studies included changes to pharmacotherapy protocols). Other studies, such as Oostlander 2019, Mangat 2019, and Ryan 2019 described randomized studies of environmental changes (D'Apolito 1999; Maichuk 1999; Oro 1988), and acupuncture (Raith 2015), in combination with a larger body of non‐randomized evidence. Based on this combination of evidence, Mangat 2019 concluded "most non‐pharmacological treatments (exception of rocking and outpatient therapy) are safe, effective and easy to be implemented as adjunct therapies to pharmacological treatments."

In summary, recent reviews of non‐pharmacological care for opioid withdrawal in newborns described the potential benefits based largely on non‐randomized evidence. In the current review of RCTs, we concluded that the evidence was uncertain. There is a need for further high‐quality studies to identify the most efficacious interventions and guide clinical care teams to choose safe, cost‐effective, and impactful non‐pharmacological care practices.

Figure 1

Study flow diagram.

Figure 2

Risk of bias summary: review authors' judgments about each risk of bias item for each included study.

Analysis 1.1

Comparison 1: Infants at risk for, or having early symptoms consistent with, opioid withdrawal, Outcome 1: Length of hospitalization (days)

Analysis 1.2

Comparison 1: Infants at risk for, or having early symptoms consistent with, opioid withdrawal, Outcome 2: Pharmacological treatment with ≥ 1 doses of opioid or sedative medication

Analysis 1.3

Comparison 1: Infants at risk for, or having early symptoms consistent with, opioid withdrawal, Outcome 3: Peak neonatal abstinence syndrome (NAS) score in first 72 hours of life

Analysis 1.4

Comparison 1: Infants at risk for, or having early symptoms consistent with, opioid withdrawal, Outcome 4: Neonatal intensive care unit admission

Analysis 1.5

Comparison 1: Infants at risk for, or having early symptoms consistent with, opioid withdrawal, Outcome 5: Days to regain birth weight

Analysis 1.6

Comparison 1: Infants at risk for, or having early symptoms consistent with, opioid withdrawal, Outcome 6: Weight nadir (% weight loss)

Summary of findings 1. Non‐pharmacological treatment compared to placebo for infants at risk for, or having early symptoms consistent with, opioid withdrawal

Outcomes	*Anticipated absolute effects (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)
Non‐pharmacological treatment compared to placebo for infants at risk for, or having early symptoms consistent with, opioid withdrawal
Patient or population: infants at risk for, or having early symptoms consistent with, opioid withdrawal Setting: newborn or neonatal unit, USA and UK Intervention: non‐pharmacological treatment Comparison: placebo
Outcomes	Risk with placebo	Risk with non‐pharmacological treatment	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)
*Primary outcome measures*
Length of hospitalization (days)
– Modifying environmental stimulation	Mean was 11.5 days (SD 3.4)	MD 1 lower (2.82 lower to 0.82 higher)	—	30 (1 RCT)	⊕⊝⊝⊝ Very low^a
– Feeding practices	—	—	—	—	—
– Support of the mother‐infant dyad	Mean was 19.4 days (SD 13)	MD 8.9 lower (19.84 lower to 2.04 higher)	—	14 (1 RCT)	⊕⊝⊝⊝ Very low^b
Pharmacological treatment with ≥ 1doses of opioid or sedative medication
– Modifying environmental stimulation	Study population		RR 1.00 (0.86 to 1.16)	92 (3 RCTs)	⊕⊕⊝⊝ Low^c
– Modifying environmental stimulation	844 per 1000 (533 to 1000)	851 per 1000 (533 to 1000)	RR 1.00 (0.86 to 1.16)	92 (3 RCTs)	⊕⊕⊝⊝ Low^c
– Feeding practices	Study population		RR 0.92 (0.63 to 1.33)	49 (1 RCT)	⊕⊝⊝⊝ Very low^d
– Feeding practices	727 per 1000	667 per 1000	RR 0.92 (0.63 to 1.33)	49 (1 RCT)	⊕⊝⊝⊝ Very low^d
– Support of the mother‐infant dyad	Study population		RR 0.50 (0.13 to 1.90)	14 (1 RCT)	⊕⊝⊝⊝ Very low^b
– Support of the mother‐infant dyad	571 per 1000	286 per 1000	RR 0.50 (0.13 to 1.90)	14 (1 RCT)	⊕⊝⊝⊝ Very low^b
Secondary outcome measures**
NICU admission
– Support of the mother‐infant dyad	Study population		RR 0.50 (0.13 to 1.90)	14 (1 RCT)	⊕⊝⊝⊝ Very low^b
– Support of the mother‐infant dyad	571 per 1000	286 per 1000	RR 0.50 (0.13 to 1.90)	14 (1 RCT)	⊕⊝⊝⊝ Very low^b
Days to regain birthweight
– Feeding practices	Mean was 14.7 days (SD 2.84)	MD 1.1 lower (2.76 lower to 0.56 higher)	—	46 (1 RCT)	⊕⊝⊝⊝ Very low^e
Weight nadir (% weight loss)
– Modifying environmental stimulation	Mean was 5.5% weight loss	MD 0.28 lower (1.15 lower to 0.59 higher)	—	194 (1 RCT)	⊕⊝⊝⊝ Very low^f
– Feeding practices	Mean was 9.4% weight loss	MD 0.8 lower (2.24 lower to 0.64 higher)	—	46 (1 RCT)	⊕⊝⊝⊝ Very low^d
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). Only interventions that reported on secondary outcomes are included in the 'Summary of findings' table. CI: confidence interval; MD: mean difference; NICU: neonatal intensive care unit; RCT: randomized controlled trial; SD:** standard deviation.
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aDowngraded three levels for very serious concerns about risk of bias (due to unblinded study, concern for performance, detection, and attrition bias) and serious concern for imprecision (due to wide CIs). ^bDowngraded three levels for very serious concerns about risk of bias (due to unblinded study, concern for performance, detection, and attrition bias) and very serious concern for imprecision (due to very wide CIs). ^cDowngraded two levels for very serious concerns about risk of bias (due to unblinded study, concern for performance, detection, and attrition bias). ^dDowngraded three levels for very serious concerns about risk of bias (due to concern for attrition and reporting bias), serious concern about imprecision (due to wide CIs), and plausible confounding that would reduce demonstrated effect. ^eDowngraded three levels for very serious concerns about risk of bias (due to concern for attrition and reporting bias), very serious concern about imprecision (due to very wide CIs), and plausible confounding that would reduce demonstrated effect. ^fDowngraded three levels for very serious concerns about risk of bias (due to unblinded study, concern for performance and detection bias) and serious concern for imprecision (due to wide CIs).

Summary of findings 1. Non‐pharmacological treatment compared to placebo for infants at risk for, or having early symptoms consistent with, opioid withdrawal

Table 1. Excluded studies of modifying environmental stimulation

Study	Details
Radziewicz 2018	Methods	Prospective cohort
	Participants	30 newborns at risk for withdrawal in a newborn nursery
	Time period	Not reported
	Site of care	Newborn nursery (Ohio, USA)
	Interventions	Quote: "Each subject received one thirty‐minute treatment of Reiki in a private room."
	Outcomes	Slight decrease in heart rate and withdrawal scores following treatment
	Notes	—
Zuzarte 2017	Methods	Prospective cohort (within‐subjects)
	Participants	26 opioid‐exposed newborns (> 37 weeks' gestation) treated pharmacologically for NAS
	Time period	Not reported
	Site of care	Neonatal intensive care unit and newborn nursery (Massachusetts, USA)
	Interventions	Quote: "A specially‐constructed mattress delivered low‐level SVS (30–60Hz, 10–12μm RMS [root‐mean‐square]), alternated in 30‐min intervals between continuous vibration (ON) and no vibration (OFF) over a 6–8 hr session."
	Outcomes	35% reduction in movement activity with SVS
	Notes	—
NAS: neonatal abstinence syndrome; SVS: stochastic vibrotactile stimulation.

Table 1. Excluded studies of modifying environmental stimulation

Table 2. Excluded studies of feeding practices

Study	Details
Abdel‐Latif 2006	Methods	Retrospective cohort
	Participants	190 consecutive drug‐dependent mother‐infant pairs
	Time period	1998–2004
	Site of care	Postnatal ward with mother unless medical or social contraindications (New South Wales, Australia)
	Interventions	Predominant feed on day 5 of life: ≤ 2 feeds daily of formula (85 infants) or > 2 feeds daily of formula (105 infants)
	Outcomes	≤ 2 feeds daily of formula vs > 2 feeds daily of formula Shorter length of hospitalization (mean: 14.7 days, SD 14.9 vs 19.1 days, SD 15.0; P = 0.049) Significant decrease in need for treatment (52.9% vs 79.0%; P < 0.001) Non‐significant decrease in treatment duration (mean: 85.4 days, SD 71.7 vs 108.2 days, SD 81.8; P = 0.185) Non‐significant decrease in maximum dose of morphine (mean: 0.57 mg/kg/day, SD 0.22 vs 0.59 mg/kg/day, SD 0.22; P = 0.526) Fewer infants in foster care (9.4% vs 29.5%; P = 0.001) Fewer infants designated; quote: "child at risk" (31.8% vs 68.6%; P < 0.001) Lower mean Finnegan scores (numeric values not provided in text) Later median time to withdrawal (10 days vs 3 days; P < 0.001)
	Notes	More heroin, non‐opioid, and polydrug use in formula group Emailed authors to clarify data
Arlettaz 2005	Methods	Retrospective cohort
	Participants	86 infants of women enrolled in a methadone maintenance program
	Time period	1996–2001
	Site of care	Unit not specified, 77% transferred to the neonatal unit for medical reasons or pharmacological treatment (Switzerland)
	Interventions	Infants who received more than half of total amount of milk during hospital stay as human milk (27 infants) vs infants who received less than half (59 infants)
	Outcomes	Quote: "Twenty‐six percent of breast‐fed babies (7/27) and 78% of formula‐fed babies (42/54) developed NAS (P < 0.01)." Pharmacological treatment rate and custody not reported by feeding status
	Notes	—
Dryden 2009	Methods	Retrospective cohort
	Participants	450 singleton pregnancies of drug misusing women prescribed substitute methadone for pregnancy 354 term substitute methadone‐exposed infants included in subgroup published in 2012
	Time period	January 2004 to December 2006
	Site of care	Postnatal ward with mother unless any specific indication for admission to the neonatal unit (UK)
	Interventions	Received maternal breast milk ≥ 72 hours (99 infants) vs formula fed or received maternal breast milk < 72 hours (351 infants)
	Outcomes	Received maternal breast milk ≥ 72 hours vs formula fed or received maternal breast milk < 72 hours Breastfeeding associated with reduced odds of requiring treatment for NAS (OR 0.55, 95% CI 0.34 to 0.88; P = 0.013) Follow‐up analysis of 354 infants with complete data (breastfed 72 infants, non‐breastfed 282 infants) showed higher median maximal weight loss in breastfed group (10.2% vs 8.5%; P = 0.003) The group of 22 exclusively breastfed infants had a median maximal weight loss of 9.3% Quote: "For infants who were not admitted to the [neonatal unit] (n=208), median age at discharge was 7 days; this did not differ between breast and formula feeders" More breastfed infants were classified as "non‐admitted" and did not require admission to the neonatal unit (52/72 [72%] vs 155/282 [55%])
	Notes	Emailed authors to clarify the relationship between the 2 publications, 2009 study is the primary study as fewer infants were included in 2012 study due to missing data.
Hodgson 2012	Methods	Retrospective cohort
	Participants	295 singleton mother‐infant dyads who received care in combined care for maternal substance use in a rooming in program
	Time period	October 2003 to December 2006
	Site of care	Postnatal; quote: "combined care nursery" with mother including those receiving pharmacotherapy (British Columbia, Canada)
	Interventions	Breastfed (196 infants) vs not breastfed (99 infants)
	Outcomes	Significant decrease in probability of an infant receiving morphine (12% of breastfed infants vs 37% of non‐breastfed infants; OR 0.21, SD 1.43, 95% CI 0.10 to 0.42; P < 0.001)
	Notes	—
Isemann 2011	Methods	Retrospective cohort
	Participants	128 infants that received pharmacotherapy for opiate withdrawal in the newborn intensive care unit (92 infants > 37 weeks, 36 infants < 37 weeks)
	Time period	January 2002 to December 2007
	Site of care	NICU (Ohio, USA)
	Interventions	Mother's breast milk ingested vs formula fed
	Outcomes	Inverse correlation between the percentage of mother's breast milk ingested and length of stay (β = –0.03; P = 0.02) Median: 12.5 days, range 3–51 days for infants who ingested maternal breast milk vs 18.5 days, range 9̫–43 days for formula‐fed infant
	Notes	—
Jansson 2008	Methods	Prospective case control
	Participants	8 breastfed methadone‐exposed infants and 8 matched formula‐feeding infants
	Time period	January 2001 to September 2005
	Site of care	Unit not specified (Maryland, USA)
	Interventions	Breastfed (8 infants) vs formula fed (8 infants)
	Outcomes	Breastfed vs formula fed Decrease in use of pharmacological treatment (12.5% vs 50%; P = 0.28) Shorter time of treatment (median: 6 days, IQR not reported vs 13.5 days, IQR 10.5) Lower day 3 NAS score (median: 4.5, IQR 3.9 vs 6.8, IQR 1.2; P non‐significant) Lower peak NAS score (median: 6.5, IQR 4.5 vs 11.0, IQR 5.0; P non‐significant)
	Notes	More medical comorbidities in breastfed group 3 breastfeeding infants not included due to failure to provide complete study specimens 1 breastfeeding infants not included due to inability to find match
Liu 2015	Methods	Retrospective cohort
	Participants	194 methadone‐maintained mother‐infant dyads > 30 weeks' gestation
	Time period	2000–2006
	Site of care	Unit not specified; quote: "two birthing units" (New South Wales, Australia)
	Interventions	Categorized in the 1st 2 days of life as predominantly breastfed (≥ 50% of feeds, 32 infants), fed expressed breast milk (≥ 15 mL expressed breast milk and breastfed < 3 times daily, 12 infants), or formula fed (≥ 50% of feeds and < 15 ml/day breast milk ingestion, 150 infants)
	Outcomes	Breastfed vs fed expressed breast milk vs formula fed No difference in percent requiring treatment for NAS (72% vs 100% vs 81%; P = 0.11) No difference in maximum dose of morphine required (median: 0.5 mg/kg/day, IQR 0.5 to 0.7 vs 0.7, IQR 0.53 to 9.84 vs 0.5, IQR 0.5 to 0.7; P = 0.20) No difference in mean NAS score (5.1, SD 1.3 vs 5.7, SD 0.9 vs 5.4, SD 1.1; P = 0.47) Quote: "Breastfeeding during the first 2 days of life was associated with a delayed onset of NAS (P = .04)"
	Notes	—
McQueen 2011	Methods	Retrospective cohort
	Participants	28 term infants exposed to methadone with symptoms of NAS
	Time period	March 2007 to March 2008
	Site of care	Unit not specified (Ontario, Canada)
	Interventions	Breastfeeding (> 75% breast milk, 8 infants) vs combination feeds (25% to 75% breast milk, 11 infants) vs formula fed (< 25% breast milk, 9 infants)
	Outcomes	Breastfed vs combination feed vs formula fed Fewer mean NAS scores on Modified Finnegan Scoring Tool (25, SD 23.5 vs 56.2, SD 39.1 vs 95.6, SD 34.6) Lower mean NAS scores on Modified Finnegan Scoring Tool (4.9, SD 2.9 vs 6.5, SD 3.7 vs 6.9, SD 4.2) Fewer infants scored > 8 (17% vs 38.8% vs 40.2%)
	Notes	More regular prenatal care in the breastfed group Emailed with authors to clarify data
Metz 2015	Methods	Retrospective cohort
	Participants	390 infants born to opioid‐dependent women undergoing comprehensive treatment during pregnancy (184 methadone, 77 buprenorphine, 129 slow‐release oral morphine) 50 infants excluded due to incomplete data
	Time period	1994 to January 2009
	Site of care	Unit not specified (Austria)
	Interventions	Breastfed infants (100 infants; methadone group 48 infants, slow‐release oral morphine group 21 infants, buprenorphine group 31 infants) vs non‐breastfed infants (250 infants; methadone group 118 infants, slow‐release oral morphine group 91 infants, buprenorphine group 41 infants)
	Outcomes	Breastfed vs non‐breastfed Decreased mean modified Finnegan score in methadone group (6.52, SD 2.50 vs 7.94, SD 2.87; P = 0.004) and slow‐release oral morphine group (6.97, SD 2.77 vs 9.07, SD 3.11; P = 0.007) with no difference in buprenorphine group (4.89, SD 2.03 vs 5.77, SD 2.49; P = 0.110) Decreased peak modified Finnegan score in methadone group (mean: 12.60, SD 5.16 vs 17.00, SD 5.65; P < 0.001) and slow‐release oral morphine group (mean 15.24, SD 6.66 vs 18.93, SD 5.56; P = 0.009) with no difference in buprenorphine group (mean: 11.19, SD 5.11 vs 12.29, SD 4.78; P = 0.352) Decreased total morphine dose in methadone group (mean: 4.35 mg, SD 9.03 vs 12.65 mg, SD 18.64; P < 0.001) with no difference in slow‐release oral morphine group (mean: 5.23 mg, SD 7.00 vs 8.75 mg, SD 11.54; P = 0.193) and buprenorphine group (mean: 1.90 mg, SD 3.32 vs 2.77 mg, SD 5.78; P = 0.457) Decreased duration of treatment in methadone group (mean: 8.10 days, SD 12.64 vs 16.94 days, SD 16.75; P < 0.001) and slow‐release oral morphine group (mean: 10.20 days, SD 12.53 vs 18.07, SD 13.67; P = 0.020) with no difference in buprenorphine group (mean: 4.74 days, SD 5.58 vs 5.76 days, SD 7.37; P = 0.524) Decreased length of stay in methadone group (mean: 17.17 days, SD 10.28 vs 29.36 days, SD 15.07; P < 0.001) and slow‐release oral morphine group (mean: 19.71 days, SD 12.37 vs 31.03 days, SD 13.81; P = 0.001) with no difference in buprenorphine group (mean: 12.26 days, SD 5.35 vs 14.71 days, SD 9.60; P = 0.174)
	Notes	Outcomes reported as subgroups only
O'Connor 2013	Methods	Retrospective cohort
	Participants	85 mother‐infant dyads in buprenorphine treatment program, infants > 35 weeks
	Time period	December 2007 to August 2012
	Site of care	Unit not specified (Maine, USA)
	Interventions	Breastfeeding (65 infants) vs non‐breastfed infants (20 infants)
	Outcomes	Breastfed vs non‐breastfed Non‐significant increase in length of hospital stay (mean: 7.08 days, SD 4.39 vs 6.60 days, SD 1.70; P = 0.35) Non‐significant decrease in peak modified Finnegan NAS score (mean: 8.83, SD 3.56 vs 9.65, SD 2.58; P = 0.17) Non‐significant decrease in pharmacological treatment (23.1% vs 30.0%; P = 0.56) Non‐significant decrease in time to first and last peak modified Finnegan NAS score (first: mean: 66.5 hours, SD 43.80 vs 73.5 hours, SD 41.82; P =.32; last: mean: 76.1 hours, SD 43.95 vs 78.3 hours, SD 38.54; P = 0.67) Decreased likelihood of higher modified Finnegan NAS scores (≥ 8: 65% vs 75%; ≥ 12: 17% vs 30%)
	Notes	Breastfed infants were significantly less likely to have exposure to illicit substances in the 3rd trimester (41.4% vs 80.0%; P = 0.004)
Pritham 2012	Methods	Retrospective cohort
	Participants	152 opioid‐dependent pregnant women in treatment (136 methadone, 16 buprenorphine) and their infants
	Time period	January 2005 to December 2007
	Site of Care	NICU (Maine, USA)
	Interventions	Breastfed (17 infants, methadone group 14, buprenorphine group 3) vs both formula and breast (26 infants, methadone group 22, buprenorphine group 4) vs formula fed (105 infants, methadone group 96, buprenorphine group 9)
	Outcomes	Length of stay was shorter in breastfed neonates than formula‐fed neonates or neonates that fed both formula and breast in regression model of methadone group (β = –0.176; P = 0.05)
	Notes	—
Radmacher 2017	Methods	Prospective cohort (historically controlled feasibility study)
	Participants	12 term formula‐fed infants receiving oral morphine for opioid withdrawal matched to historical controls
	Time period	2014–2017
	Site of care	NICU (Kentucky, USA)
	Interventions	Exclusive donor milk for up to 2 weeks (12 infants) vs formula‐fed historical controls (12 infants)
	Outcomes	Donor milk vs formula fed No significant difference in length of stay (mean: 41 days, SD 16 vs 33 days, SD 9; P non‐significant) No significant difference in length of treatment (mean: 37 days, SD 16 vs 29 days, SD 9; P non‐significant) No significant difference in median total Finnegan Neonatal Abstinence Score (values not reported) Significant decrease in gastrointestinal subscore > 2 (39% vs 61%; P = 0.001) No significant difference in return to birth weight (mean: 17 days, SD 7 vs 16 days, SD 4; P non‐significant) No significant difference in weight gain during study period (mean: 19.1 g/day, SD 13.1 vs 20.5 g/day, SD 3.0; P non‐significant)
	Notes	Benzodiazepine exposure in the donor milk group but not the control group Significant deviation from intended intervention noted
Short 2016	Methods	Retrospective cohort
	Participants	3725 singleton in‐hospital births > 35 weeks' gestation diagnosed with NAS based on discharge diagnosis codes
	Time period	2012–2014
	Site of care	Unit not specified, though mentions benefits of rooming in and 40% admission to NICU (Pennsylvania, USA)
	Interventions	Breastfeeding at discharge (1576 infants) vs not breastfeeding (1968 infants) as reported on birth certificate
	Outcomes	Length of stay was 9.4% shorter in infants breastfed at discharge than infants not breastfed at discharge in regression model (β = –0.060; P = 005)
	Notes	No information about in‐utero exposure or postnatal treatment for NAS Breastfed infants were more likely to have a higher birthweight and gestational age; less likely to be insured by Medicaid; and more likely to be born to mothers who completed high school, were married, had adequate prenatal care, and who did not smoke
Wachman 2013	Methods	Prospective cohort
	Participants	86 infants > 36 weeks' gestation exposed to methadone or buprenorphine in utero
	Time period	July 2011 to July 2012
	Site of care	Unit not specified, 5 tertiary care centers and community hospitals (Massachusetts and Maine, USA)
	Interventions	Assessed breastfeeding as a potential covariate related to NAS severity; breastfeeding (38 infants) vs no breastfeeding (48 infants)
	Outcomes	Breastfed vs non‐breastfed Shorter length of stay (mean: 15.8 days, 95% CI 11.5 to 20.1 vs 27.4 days, 95% CI 22.5 to 32.3; P < 0.001) Decrease in pharmacological treatment (50% vs 77%; P = 0.009)
	Notes	Objective: to determine whether single nucleotide polymorphisms in the μ‐opioid receptor (OPRM1), multiple‐drug resistance (ABCB1), and catechol‐O‐methyltransferase (COMT) genes are associated with length of hospital stay and the need for treatment of NAS
Welle‐Strand 2013	Methods	Retrospective and prospective cohort
	Participants	124 infants of women on opioid maintenance therapy (78 methadone, 46 muprenorphine)
	Time period	1999–2003 (36 infants, retrospective), 2005–2007 (36 infants, prospective), 2007–2009 (52 infants, retrospective)
	Site of care	Unit not specified, national cohort (Norway)
	Interventions	Breastfed (95 infants; methadone exposed 58, buprenorphine exposed 37) vs not breastfed (29 infants; methadone exposed 20, buprenorphine exposed 9)
	Outcomes	Breastfed vs not breastfed Non‐significant decrease in NAS use of pharmacological treatment overall (overall: 57% vs 69%, not significant; methadone exposed: 53% vs 80%; P < 0.05; buprenorphine exposed: 64% vs 44%; P not significant) Significant decrease in NAS treatment duration overall (overall: mean: 28.6 days, SD 19.1 vs 46.7 days, SD 26.3; P < 0.05; methadone‐exposed mean: 31.0 days, SD 21.4 vs 48.9 days, SD 27.2; P < 0.05; buprenorphine‐exposed mean: 25.7 days, SD 16.0 vs 38.8 days, SD 24.0; P not significant) Length of NAS treatment was shorter in breastfed infants vs not breastfed infants in regression model (overall: β = –19.7; P = 0.00; methadone exposed: β = –20.7; P = 0.00; buprenorphine exposed: β = –6.62; P = 0.59)
	Notes	More preterm birth and delivery by cesarean in non‐breastfed group
CI: confidence interval; NAS: neonatal abstinence syndrome; IQR: interquartile range; NICU: neonatal intensive care unit; OR: odds ratio; SD: standard deviation.

Table 2. Excluded studies of feeding practices

Table 3. Excluded studies of support of the mother‐infant dyad

Study	Details
Abrahams 2007	Methods	Retrospective cohort
	Participants	106 infants of mothers with known heroin or methadone use during pregnancy
	Time period	October 2001 to December 2002
	Site of care	Postnatal ward and intensive care nursery (BC, Canada)
	Interventions	Rooming in infants in study hospital (32 infants) compared to historical comparisons in study hospital (38 infants) and comparisons at a neighboring hospital without rooming in (36 infants)
	Outcomes	Rooming in vs historical comparisons in study hospital vs neighboring hospital without rooming in Shorter length of hospital stay (mean: 11.8 days, SD 9.1 vs 23.5 days, SD 24.6 vs 25.9 days, SD 19.7) Statistically significant decrease in use of pharmacological treatment (25.0% vs 55.3% [RR 0.45, 95% CI 0.23 to 0.87] vs 52.8% [RR 0.47, 95% CI 0.24 to 0.93]) Statistically significant decrease in pharmacological treatment duration (mean: 5.9 days, SD 14.2 in study hospital vs 18.6 days, SD 23.4 vs 18.6 days, SD 20.1) Statistically significant increase in likelihood of discharge with mothers (discharged in custody of mothers: 71.9% vs 31.6% [RR 2.28, 95% CI 1.36 to 3.81] vs 42.5% [RR 1.52, 95% CI 1.01 to 2.29]) Trend toward decreased prevalence of NAS symptoms Increased prevalence of weight loss > 10% in the first week of life (16.6% vs 5.2% [RR 2.96, 95% CI 0.62 to 14.28] vs 5.3% [RR 2.81, 95% CI 0.59 to 13.5]) Statistically significant decrease in admission to a level II nursery (37.5% vs 89.5% [RR 0.42, 95% CI 0.26 to 0.66] and 83.3% ]RR 0.45, 95% CI 0.28 to 0.72])
	Notes	Several baseline differences between groups and hospitals with significant difference in feeding practices between groups (62.5% of the rooming in group breastfeeding compared to 7.9% of the historical cohort and 11.1% of the concurrent cohort)
Abrahams 2010	Methods	Retrospective cohort
	Participants	952 singleton infants of mothers with documented substance use
	Time period	October 2003 to December 2006
	Site of care	Postnatal; quote: "combined care nursery" and intensive care nursery (British Columbia, Canada)
	Interventions	Rooming in with comprehensive education in study hospital (355 infants, term 255 infants) vs standard nursery care at all other hospitals (597 infants, term 434 infants)
	Outcomes	Rooming in vs standard nursery care Longer hospital length of stay (all infants: mean: 20.7 days, SD 20.5 vs 10.7 days, SD 19.4; term infants: mean 17.6 days, SD 11.3 vs 8.1 days, SD 12.6) Decreased admissions to NICU (all infants: 38.9% vs 45.0%; OR 0.68, 95% CI 0.51 to 0.92; P < 0.001; term infants: 23.5% vs 38.5%; OR 47, 95% CI 0.324 to 0.67; P < 0.001) Shorter NICU length of stay in term infants (all infants: mean: 5.0 days, SD 19.2 vs 5.0 days, SD 12.6; term infants: mean: 1.1 days, SD 3.1 vs 3.1 days, SD 8.3) Increased receipt of breast milk (all infants: 63.7% vs 45.4%; OR 2.11, 95% CI 1.61 to 2.77; P < 0.001) No difference in breastfeeding status at discharge (all infants: 34% vs 35.1%; OR 0.953, 95% CI 0.67 to 1.35; P = 0.79) Increased discharge home with mother (69.9% vs 58.7%)
	Notes	Baseline differences between groups and hospitals. Newborns in rooming in group had increased odds of receiving breast milk during hospital stay.
Crook 2017	Methods	Retrospective and prospective cohort (quality improvement methodology)
	Participants	200 infants evaluated for NAS
	Time period	February 2014 to July 2015
	Site of care	NICU, small percentage on pediatric unit if high NICU census (North Carolina, USA)
	Interventions	Baby friendly status Breastfeeding initiative (3 class curriculum)
	Outcomes	Decrease in length of stay from a mean of 18.80 days, SD 14.71 to 10.41 days, SD 10.54 (P < 0.001) Decrease in pharmacological treatment from 67.3% to 34.8% (P < 0.001) Increase in exclusive breastfeeding from 20.0% to 31.9% (P not reported)
	Notes	Quote: "Infants who received more than 50% human milk had a predicted [length of stay] 2.78 days shorter than infants who received more than 50% formula during hospitalization (P = .08)."
Howard 2017	Methods	Retrospective cohort
	Participants	86 mother‐infant dyads in a single center rooming in cohort of infants treated pharmacologically for NAS
	Time period	March 2015 to April 2016
	Site of care	Pediatric inpatient unit for pharmacological treatment (Massachusetts, USA)
	Interventions	Parental presence (percent of time)
	Outcomes	Maximum (100%) parental presence was associated with a 9‐day shorter length of stay (r = –0.31, 95% CI –0.48 to –0.10; P = 0.01), 8 fewer days of infant opioid therapy (r = –0.34, 95% CI –0.52 to –0.15; P = 0.001), and 1 point lower mean Finnegan score (r = –0.35, 95% CI –0.52 to –0.15; P = 0.01) Quote: "There was a 5.3 mg decrease in the total morphine equivalent dose with increased parental presence (r = –0.20; 95% CI, –0.39 to 0.02; P = .06)" Quote: "The mean NAS score when a parent was present was significantly lower compared with when a parent was not present (5.1 [95% CI, 4.9–5.3] versus 6.0 [95% CI, 5.8–6.2]; P < .0001)" Quote: "Parental presence was higher for infants who were breastfed (65.2% vs 44.5%; P < .001)" Quote: "Parental presence was lower for infants in Department of Children and Families (DCF) custody (36.6% vs 59.7%; P < .001)" Quote: "After adjusting for breastfeeding, parental presence remained significantly associated with reduced NAS score and opioid treatment days"
	Notes	Quote: "In unadjusted analyses, any amount of breastfeeding was associated with a decreased [length of stay] (16.5 vs 21.2 days; P < .01) as was clinical trial participation (15.8 vs 20.0 days; P < .01). Similarly, breastfeeding and clinical trial participation were also significantly associated with decreased duration of opioid therapy and decreased total morphine equivalent dose."
Hünseler 2013	Methods	Retrospective cohort
	Participants	77 infants with NAS transferred from the maternity ward to the neonatal ward for potential pharmacological treatment
	Time period	2004–2011
	Site of care	Neonatal ward (Germany)
	Interventions	Rooming in within the neonatal ward (24 infants) vs not rooming in within the neonatal ward (53 infants)
	Outcomes	Rooming in vs no rooming in Shorter length of hospital stay (mean: 36.6, SD 10.2 vs 42.8, SD 15.3; P = 0.077) No difference in use of pharmacological treatment (79% vs 88.7%; P = 0.14) Shorter duration of pharmacological treatment (mean: 29.5, SD 11.5 vs 35.8, SD 15.3; P = 0.043) Lower maximum dose of opioid (mean: 132 μg/kg/day diluted tincture of opium, SD 56 vs 145 μg/kg/day diluted tincture of opium, SD 60; P = 0.84) No difference in the frequency of Finnegan Scores ≥ 12 (6.4% vs 6.3%; P = 0.98). No difference in maximum Finnegan score (19 points in both groups). Lower costs (mean EUR 10,620, SD 5306 vs EUR 14,331, SD 6028; P = 0.014) More infants discharged home with families (79.2% vs 69.8%; P = 0.046)
	Notes	Only 2 infants were breastfed
McKnight 2016	Methods	Retrospective cohort (historical controls)
	Participants	44 singleton infants > 36 weeks' gestation of women taking chronic opioid therapy
	Time period	May 2012 to September 2014 (excluding June to August 2013)
	Site of care	Pediatrics unit with transfer to NICU for pharmacological treatment (Ontario, Canada)
	Interventions	Baseline (24 infants) and postimplementation (20 infants) of initial observation while rooming in with mother in a private room in the pediatrics unit. Infants transferred to the NICU if pharmacological treatment was required
	Outcomes	Shorter length of hospitalization (median: 5.0, IQR 3 vs 24.0, IQR 34; P < 0.001) Reduced use of pharmacological treatment of NAS (15.0% vs 83.3%; P < 0.001) No difference in duration of treatment with morphine (median: 24.0, IQR could not be calculated vs 29.5, IQR 23; P = 0.83) No difference in maximum daily dose of morphine (median: 0.48 mg/kg/day, IQR could not be calculated vs 0.48 mg/kg/day, IQR 0.48; P = 0.90) Fewer treated with adjunct pharmacotherapy (0% vs 30.0%; P = 0.54)
	Notes	Quote: "While not significant, a higher proportion of infants in the rooming‐in group were breastfed exclusively or in combination with formula." Emailed authors to clarify data and the relationship between the 2 publications, 1 additional participant included in earlier publication (Newman 2015) who was different from the usual population
Metz 2011	Methods	Retrospective cohort
	Participants	114 opioid‐exposed newborns of mothers maintained on methadone (70 infants) or buprenorphine (44 infants)
	Time period	June 2005 to December 2009
	Site of care	Unit not specified (Austria)
	Interventions	Participation in a randomized double‐blind double‐dummy clinical trial ("MOTHER study", 37 infants) vs standard care (77 infants) Clinical trial included daily maternal contact with study staff and mandatory rooming in with mother for ≥ 10 days after delivery Standard care neonates were treated in a separate ward and mothers were discharged when medically cleared
	Outcomes	Clinical trial vs standard protocol Shorter length of stay (days) in the clinical trial group (methadone‐exposed mean: 16.74, SD 6.78 vs 29.36, SD 17.94; buprenorphine‐exposed mean: 13.67, SD 2.63 vs 13.92, SD 7.33) More infants used pharmacological treatment in the clinical trial group (76% vs 60%) Lower total medication dose in the clinical trial group (methadone‐exposed mean: 5.18 mg, SD 6.94 vs 21.61 mg, SD 26.64; buprenorphine‐exposed mean: 2.02 mg, SD 1.71 vs 4.30 mg, SD 7.30) Shorter length of pharmacological treatment in the clinical trial group (methadone‐exposed mean: 9.53 days, SD 9.14 vs 21.25 days, SD 21.22; buprenorphine‐exposed mean: 7.33 days, SD 4.46 vs 6.62 days, SD 8.07)
	Notes	Modified Finnegan used for clinical trial vs traditional Finnegan for standard care Likely fundamentally different group self‐selected for clinical trial
Ordean 2015	Methods	Retrospective cohort
	Participants	94 pregnant methadone‐maintained women attending integrated care programs and their children
	Time period	1997–2009
	Site of care	Unit not specified for rooming in cohort, NICU (Ontario, British Columbia, and Quebec, Canada)
	Interventions	Rooming in approach (36 infants) vs 2 neighboring hospitals that admit infants to the neonatal intensive care unit for NAS scoring (36 and 22 infants)
	Outcomes	Rooming in vs hospital 1 (no rooming in) vs hospital 2 (no rooming in) Longer hospital stay (mean: 26 days, SD 16 vs 14 days, SD 15 vs 16 days, SD 14) No difference in use of pharmacological treatment (29% vs 31% vs 18%; P = 0.564) No difference in duration of NAS treatment (mean: 13 days, SD 6 vs 25 days, SD 17 vs 30 days, SD 27; P = 0.086) Decreased NICU admission (42% vs 94% vs 91%; P = 0.003) No difference in weight loss (7.4% vs 7.0% vs 8.7%; P = 0.204) No difference in days to weight gain (mean: 4.3, SD 1.9 vs 4.2, SD 2.5 vs 5.6, SD 2.8; P = 0.093) Increase in discharge with mother (41% vs 69% vs 68%; P = 0.115) No difference in breastfeeding (14% vs 17% vs 22%; P = 0.767)
	Notes	Minimal difference in breastfeeding between groups. Significant intersite differences were found in ethnicity and marital status. Emailed authors to clarify data.
Saiki 2010	Methods	Retrospective cohort (historical controls)
	Participants	60 infants with NAS
	Time Period	2002–2007
	Site of Care	Postnatal ward for monitoring and pharmacological treatment and neonatal unit (UK)
	Interventions	Baseline (42 infants) and postimplementation (18 infants) of a system of pharmacological treatment of the infant in the postnatal ward instead of transferring from the postnatal ward to the neonatal unit if escalating symptoms
	Outcomes	Baseline vs postimplementation Shorter hospital stay (mean: 19.8 days vs 15.9 days; P = 0.012) Fewer infants required treatment (45% vs 11%; P = 0.012) Shorter duration of treatment (mean: 12.7 days vs 7.3 days; P = 0.05) No difference in discharge with mother (60% vs 67%; P = 0.264)
	Notes	—
Summey 2018	Methods	Retrospective cohort
	Participants	466 infants receiving care for NAS
	Time period	2006–2014
	Site of care	Postnatal ward for monitoring and pharmacological treatment and Neonatal unit (South Carolina, USA)
	Interventions	Infants enrolled in the Managing Abstinence in Newborns (MAiN) intervention (110 infants) vs infants who were potentially eligible but did not receive the intervention (356 infants) MAiN; quote: "provides multidisciplinary, coordinated, community‐based care for infants at risk for opioid withdrawal and their families within a mother/baby unit. MAiN's three essential elements are early treatment (methadone administered within 48 hours of birth), the option to room in with the mother for the entire hospitalization, and a combined inpatient/outpatient wean."
	Outcomes	MAiN vs no MAiN Non‐significant decrease in length of stay (median: 8 days, IQR 6 to 10 vs 9 days, IQR 4 to 18; P = 0.14) Significantly higher involvement of child protective services in the MAiN cohort (32.7% vs 23.3%; P = 0.048) All infants received level 1 care, 68.8% of comparison babies received level 2 to 3 care Lower total cost (median: USD 10,058, IQR USD 7935 to USD 11,518 vs USD 18,262, IQR USD 5816 to USD 40,922; P < 0.001) 'Convulsions' were reported in 1 intervention infant and no comparison infants Diagnosed with 'abnormal loss of weight' more often (3.6% vs 0.8%; P = 0.04)
	Notes	No significant difference in breastfeeding between groups (43.6% vs 48.9%; P = 0.57)
CI: confidence interval; DCF: Department of Children and Families; IQR: interquartile range; MAiN: Managing Abstinence in Newborns; MOTHER: Maternal Opioid Treatment: Human Experimental Research; NAS: neonatal abstinence syndrome; NICU: neonatal intensive care unit; OR: odds ratio; r: Pearson's correlation coefficient; RR: risk ratio; SD: standard deviation.

Table 3. Excluded studies of support of the mother‐infant dyad

Table 4. Excluded studies of multiple non‐pharmacological care practices

Study	Details
Grossman 2017	Methods	Retrospective and prospective cohort (quality improvement methodology)
	Participants	287 infants diagnosed with NAS. Excluded infants not exposed to methadone and with serious comorbid conditions
	Time period	January 2008 to June 2016
	Site of care	Pediatrics unit, postnatal ward, and NICU (Connecticut, USA)
	Interventions	Quote: "Standardization of nonpharmacologic care coupled with an empowering message to parents, development of a novel approach to assessment, administration of morphine on an as‐needed basis, and transfer of infants directly to the inpatient unit." Baseline period from January 2008 to February 2010 (55 infants), intervention period (188 infants), and postimplementation period from May 2015 to June 2016 (44 infants)
	Outcomes	Decrease in mean length of stay (22.4 days to 5.9 days; P < 0.001), postnatal morphine treatment (98% to 14%; P < 0.001), and mean cost (USD 44,824 to USD 10,289; P < 0.001) More infants fed breast milk at discharge following implementation (20% to 45%; P = 0.01) Fewer infants admitted to the NICU (100% to 20%; P < 0.001)
	Notes	—
Holmes 2016	Methods	Retrospective and prospective cohort (quality improvement methodology)
	Participants	163 opioid‐exposed newborns Excluded infants with gestational age < 35 weeks, another reason for NICU admission, or who completed treatment at another facility
	Time period	March 2012 to February 2015
	Site of care	Pediatrics unit, postnatal ward, and NICU (New Hampshire, USA)
	Interventions	Staff training, family education (involvement in symptom monitoring and non‐pharmacological treatment), and care in inpatient pediatric unit Baseline period from March 2012 to February 2013 (54 infants), intervention year 1 from March 2013 to February 2014 (61 infants), and intervention year 2 from March 2014 to February 2015 (48 infants)
	Outcomes	Decrease in use of pharmacological treatment (46% to 27%), adjunctive use of phenobarbital (13% to 2%), mean length of hospitalization for pharmacologically treated infants (16.9 days to 12.3 days), mean cost per treated infant (USD 19,737 to USD 8755; P < 0.01), mean cost per at risk patient (USD 11,000 to USD 5300; P < 0.01), and cumulative morphine exposure (13.7 mg to 6.6 mg per treated newborn) No change noted in length of hospitalization for non‐pharmacologically treated infants (4.2 days to 4.4 days; P = 0.33) No change in mean daily Finnegan score; quote: "no significant difference in median score, maximum score, or first score by year (P = 0.53, 0.29, 0.48, respectively)" No change in number of newborns discharged in parental custody (93% to 90%; P = 0.73)
	Notes	—
Kirchner 2014	Methods	Retrospective cohort
	Participants	131 infants who completed an international multicenter clinical trial ("MOTHER study")
	Time period	2005–2008
	Site of care	Unit not specified (international)
	Interventions	Comparison by site of enrollment; quote: "the 7 contributing sites were divided into 3 clusters: the European cluster (Vienna site, n=37), the rural US cluster (N=39): Burlington (University of Vermont, n=26), Nashville (Vanderbilt University, n=13), and the urban US cluster (n=55): Detroit (Wayne State University, n=12), Providence (Brown University, n=3), Philadelphia (Thomas Jefferson University, n=23), Baltimore (Johns Hopkins University, n=17)" Quote: "A standardized NAS rating and treatment protocol was applied, while non‐pharmacologic care of NAS symptoms differed across sites" Quote: "In Europe, all neonates (also those without treatment requirement) stayed in the hospital accommodated in a room with their mothers for at least 10 days" Quote: "The breastfeeding rate was much higher in Europe and was highest in rural US"
	Outcomes	Quote: "The lowest morphine doses for both medication conditions were administered in Europe where all babies were kept in a rooming‐in situation in hospital for at least 10 days" Pharmacological treatment rates: 18.5% Europe rooming in model vs 29.5% urban USA vs 18.5% rural USA Mean total dose of morphine within the first 28 postpartum days: 5.38 mg Europe rooming in model vs 35.05 mg urban USA vs 8.66 mg rural US Quote: "Urban US neonates had significantly higher NAS scores (p<0.01) compared to rural US and European neonates, and needed significantly higher morphine doses (P < 0.05) with longer treatment duration" Quote: "Urban US neonates exhibited most neurological symptoms (P < 0.001) while in Europe autonomous, respiratory and gastrointestinal symptoms were found significantly more often compared to urban and/or rural US (P < 0.05)"
	Notes	Quote: "No differences between sites could be found in behavioural symptoms, but interestingly a significant positive influence on these symptoms could be demonstrated by breastfeeding (β = –0.261; P = 0.034), with explaining 6.8% of variance"
Loudin 2017	Methods	Retrospective cohort (historical controls)
	Participants	1023; quote: "inborn neonates >35 weeks' gestational age with the diagnosis of NAS (ICD9‐CM 779.5), requiring pharmacologic treatment"
	Time period	2010–2015
	Site of care	Neonatal abstinence center, NTU, and NICU (West Virginia, USA)
	Interventions	Baseline (number of infants not reported); quote: "managed in the NICU, a 36 bed level III unit with an average of 550 admissions per year" and post‐implementation (number of infants not reported) of a neonatal therapeutic unit (NTU), (quote) "a 15‐bed low‐light, low‐noise unit with a dedicated nursing staff and nursing aids along with community volunteers called 'rockers' who were specially trained to hold and rock the neonates when appropriate and when family members were unavailable. The staff utilized therapeutic handling techniques that further enhanced calming effects on the neonates." Also compared to post‐implementation (78 infants) of an offsite; quote: "neonatal abstinence center with 12 beds and a dedicated staff of nurses, patient care assistants, and volunteers trained in the care of neonates with NAS. In addition, on‐site social service personnel served to transition caretakers (birth parents, foster parents or adopting families) … to a safe home environment. Limited number of patient rooms were available to accommodate rooming in, thus that space was utilized to help prepare caretakers prior to discharge."
	Outcomes	Quote: "The median [length of stay] for NAS medication‐treated patients in the NICU without comorbidities was 24 days (interquartile range 24 to 52). Median [length of stay] in the NTU was 26 days (interquartile range 26 to 42). The median [length of stay] for Lily's Place was 33 days (interquartile range 32 to 60 days), indicative of the transition to home life mission of the facility in addition to extended neonatal withdrawal care" Significantly higher median cost per patient for those managed in the NICU (USD 90,601, IQR USD 64,489 to USD 128,135) compared to the NTU (USD 68,750, IQR USD 44,952 to USD 92,548) and then compared to the offsite neonatal abstinence center (USD 17,688, IQR USD 9933 to USD 20,033); P < 0.0001 Quote: "The percentage of total NAS patients requiring pharmacologic treatment in the NICU fell from 100% prior to the establishment of the NTU to 22% by 2015"
	Notes	—
Miles 2007	Methods	Retrospective cohort (historical controls)
	Participants	176 infants of women in a methadone maintenance program
	Time period	1991–1994, 1997–2001
	Site of care	Postnatal ward and neonatal medicine unit (UK)
	Interventions	Baseline (78 infants) and postimplementation (98 infants) of a shared care approach in clinical management with a drug liaison midwife service for mothers and infants. Also changed from primary admission to the neonatal medicine unit to primary admission to the maternity ward with transfer to the neonatal medicine unit if pharmacological treatment was required. Additionally, scoring was modified, the threshold for pharmacological treatment adjusted, and treatment choice changed from chlorpromazine to phenobarbitone. Quote: "The [drug liaison midwife] provided specialist advice regarding methadone treatment, care of the newborn, and the advantages of breastfeeding."
	Outcomes	Quote: "Infants spend less time in hospital (median 5 days vs 28 days, P < 0.0002), a smaller proportion had treatment for NAS (14% vs 79%, P = 0.001), and neonatal medicine unit admission was reduced (median 14 days vs 26 days, P < 0.0003 [number of infants 100% vs 40%, P < 0.001])" Quote: "Neonatal convulsions (P = 0.0001) and jaundice (P < 0.001) occurred less frequently, and more infants were breastfed (P = 0.001)" Quote: "About 10% of infants were discharged to a foster family" in both time periods Quote: "None of these children had impairments of vision, hearing, growth, or development" in the first year after leaving the hospital (baseline n = 24, post‐implementation n = 98)
	Notes	—
Patrick 2016	Methods	Retrospective and prospective cohort (quality improvement methodology)
	Participants	3458 infants pharmacologically treated for NAS. Data were collected through serial cross‐sectional audits of 199 participating centers
	Time period	2012–2014
	Site of care	Unit not specified (International)
	Interventions	Quote: "Participating NICUs shared common interests, worked collaboratively, shared information and ideas, and developed standardized practices." The goal was to facilitate guideline update "for the following: (1) screening for maternal substance use; (2) nonpharmacologic treatment of infants with NAS; (3) scoring signs of NAS; (4) breastfeeding; (5) pharmacologic management; and (6) duration of observation of exposed infants."
	Outcomes	Quote: "(T)he mean number of NAS‐focused guidelines increased from 3.7 to 5.1 of a possible 6 (P < .001), with improvements noted in all measured domains" Quote: "(D)ecreases occurred in median (interquartile range) length of pharmacologic treatment, from 16 days (10 to 27 days) to 15 days (10 to 24 days; P = .02) and [length of stay] from 21 days (14 to 33 days) to 19 days (15 to 28 days; P = .002)" The difference in predicted mean length of treatment for hospitals that implemented policies about non‐pharmacological treatment of NAS was –2.0 days (95% CI –4.4 to 0.5) with a predicted length of stay of –2.5 days (95% CI –5.1 to 1.2) The difference in predicted mean length of treatment for hospitals that implemented policies about breastfeeding or provision of expressed human milk was –0.7 days (95% CI –2.1 to 0.7) with a predicted length of stay of –0.5 days (95% CI –2.1 to 1.1)
	Notes	—
Wachman 2018	Methods	Retrospective and prospective cohort (pre‐/postquality improvement methodology)
	Participants	240 opioid‐exposed infants > 36 weeks' gestation
	Time period	April 2015 to December 2017
	Site of care	Pediatrics unit and postnatal ward (Massachusetts, USA)
	Interventions	Baseline (101 infants), intervention (54 infants), and postintervention (85 infants). Interventions included (quote): "a non‐pharmacologic care bundle, function‐based assessments consisting of symptom prioritization and then the "Eat, Sleep, Console" (ESC) Tool; and a switch to methadone for pharmacologic treatment."
	Outcomes	Quote: "Pharmacologic treatment decreased from 87.1 to 40.0%; adjunctive agent use from 33.6 to 2.4%; hospitalization length from a mean 17.4 to 11.3 days, and opioid treatment days from 16.2 to 12.7 (P < 0.001 for all). Total hospital charges decreased from $31,825 to $20,668 per infant. Parental presence increased from 55.6 to 75.8% (P < 0.0001). No adverse events were noted." Hospital length of stay for pharmacologically treated infants trended downward (19.1 days to 17.6 days; P = 0.11) Quote: "There were no NICU admissions for management of NAS or its complications, and no seizures" No significant change in NICU admission (23.7% to 21.2%; P = 0.78) or DCF custody status (19.8% to 28.2%; P = 0.18)
	Notes	Quote: "Breastfeeding initiation rate did not differ during the two time periods."
CI: confidence interval; DCF: Department of Children and Families; ESC: Eat, Sleep, Console Tool; IQR: interquartile range; MOTHER: Maternal Opioid Treatment: Human Experimental Research; NAS: neonatal abstinence syndrome; NICU: neonatal intensive care unit; NTU: neonatal therapeutic unit.

Table 4. Excluded studies of multiple non‐pharmacological care practices

Comparison 1. Infants at risk for, or having early symptoms consistent with, opioid withdrawal

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Length of hospitalization (days) Show forest plot	2		Mean Difference (IV, Fixed, 95% CI)	Subtotals only

1.1.1 Modifying environmental stimulation	1	30	Mean Difference (IV, Fixed, 95% CI)	‐1.00 [‐2.82, 0.82]
1.1.2 Feeding practices	0	0	Mean Difference (IV, Fixed, 95% CI)	Not estimable
1.1.3 Support of the mother‐infant dyad	1	14	Mean Difference (IV, Fixed, 95% CI)	‐8.90 [‐19.84, 2.04]
1.2 Pharmacological treatment with ≥ 1 doses of opioid or sedative medication Show forest plot	5		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only

1.2.1 Modifying environmental stimulation	3	92	Risk Ratio (M‐H, Fixed, 95% CI)	1.00 [0.86, 1.16]
1.2.2 Feeding practices	1	49	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.63, 1.33]
1.2.3 Support of the mother‐infant dyad	1	14	Risk Ratio (M‐H, Fixed, 95% CI)	0.50 [0.13, 1.90]
1.3 Peak neonatal abstinence syndrome (NAS) score in first 72 hours of life Show forest plot	3		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

1.3.1 Modifying environmental stimulation	3		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.3.2 Feeding practices	0		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.3.3 Support of the mother‐infant dyad	0		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.4 Neonatal intensive care unit admission Show forest plot	1	14	Risk Ratio (M‐H, Fixed, 95% CI)	0.50 [0.13, 1.90]

1.4.1 Modifying environmental stimulation	0	0	Risk Ratio (M‐H, Fixed, 95% CI)	Not estimable
1.4.2 Feeding practices	0	0	Risk Ratio (M‐H, Fixed, 95% CI)	Not estimable
1.4.3 Support of the mother‐infant dyad	1	14	Risk Ratio (M‐H, Fixed, 95% CI)	0.50 [0.13, 1.90]
1.5 Days to regain birth weight Show forest plot	1	46	Mean Difference (IV, Fixed, 95% CI)	‐1.10 [‐2.76, 0.56]

1.5.1 Modifying environmental stimulation	0	0	Mean Difference (IV, Fixed, 95% CI)	Not estimable
1.5.2 Feeding practices	1	46	Mean Difference (IV, Fixed, 95% CI)	‐1.10 [‐2.76, 0.56]
1.5.3 Support of the mother‐infant dyad	0	0	Mean Difference (IV, Fixed, 95% CI)	Not estimable
1.6 Weight nadir (% weight loss) Show forest plot	2		Mean Difference (IV, Fixed, 95% CI)	Subtotals only

1.6.1 Modifying environmental stimulation	1	194	Mean Difference (IV, Fixed, 95% CI)	‐0.28 [‐1.15, 0.59]
1.6.2 Feeding practices	1	46	Mean Difference (IV, Fixed, 95% CI)	‐0.80 [‐2.24, 0.64]
1.6.3 Support of the mother‐infant dyad	0	0	Mean Difference (IV, Fixed, 95% CI)	Not estimable

Comparison 1. Infants at risk for, or having early symptoms consistent with, opioid withdrawal