Early Weight Loss Percentile Curves and Feeding Practices in Opioid-Exposed Infants

In this retrospective observational cohort study, we examined data from patient charts of infants with NOWS managed by the ESC approach from January 2014 to July 2019 at Yale New Haven Children’s Hospital. The ESC approach encourages rooming-in and intensive first-line nonpharmacologic interventions. Decisions on initiation, escalation, and reduction of pharmacologic treatment is based on provider assessment of 3 patient factors: eating, sleeping, and consolability.⁹  During the study period, if pharmacologic treatment was required, infants received either morphine and clonidine or morphine alone. Before April 2015, infants were placed on a weaning protocol that decreased the morphine dose by 10% of the peak morphine dose daily. From April 2015 onward, medications were used as needed and not scheduled or weaned. Infants were managed in the newborn nursery and transferred to the general inpatient unit when mothers were discharged from obstetrician care. They were only transferred to the NICU if they had an additional medical problem requiring intensive care or if a bed was not available on the pediatric inpatient unit.

The study population included infants born with prenatal exposure to opioids who were managed on the general inpatient unit of a tertiary academic medical center with ∼ 4500 births annually. Diagnosis of exposure was made with maternal history and maternal urine toxicology results at time of delivery. Subjects were identified with administrative codes for NOWS and prenatal opioid exposure (International Classification of Diseases, Ninth Revision codes 779.5 and 760.72 or Tenth Revision codes P96.1 and P04.49) and confirmed by chart review, at which point data of interest were obtained. Infants were given these diagnoses when they were prenatally exposed to opioids and presented any signs of withdrawal. We excluded infants born <35 weeks’ gestation, infants with prenatal exposure to opioids <30 days, and/or those who were transferred to the NICU at any point of their birth hospital stay. Infants were transferred to the NICU if they had an additional problem requiring medical care or if there were no beds available on the pediatric inpatient unit (Supplemental Table 3).

We abstracted daily weight (grams) from birth to either discharge or day 7 of life, whichever was earlier. Data were recorded using whole hours for every day. Our primary outcome, daily percent weight loss from birth weight (BW), was calculated as daily weight minus BW divided by the BW and multiplied by 100. Secondary outcomes included maximum percent weight loss, weight nadir (calculated as the day of life of maximum percent weight loss), length of stay (LOS), percent weight loss at 48 hours and 72 hours, receipt of increased caloric density feeds (considered as 22 kcal/oz or above), requirement for nasogastric tube (NGT) feedings, and occurrences of >10% weight loss. Increased caloric density feeds were achieved with fortified Similac powder or breast milk fortified with Similac powder to 22, 24, or 27 kcal/oz.

Newborn and maternal characteristics, newborn outcomes, and characteristics of infants receiving NGT feedings were stratified by vaginal and cesarean delivery types and summarized using medians (interquartile range [IQR]) and count (percent). Using mixed-effects quantile regression modeling, we estimated 50^th, 75^th, and 90^th percentiles of weight loss (percent weight loss from BW) across the first 168 hours of life^37–41  stratified by delivery type. This regression approach uses direct maximum likelihood estimation, in which the model parameters are estimated in a way that maximizes the probability of the observed data across all infants regardless of length of stay. Therefore, missing outcomes are assumed to be missing at random (MAR),^40,42  whereby the estimated weight percentile across each of 7 days takes into account contributions from all infants because the likelihood of all observed data are being maximized for infants for varying lengths at the same time. To further minimize the potential impact of missing outcomes because of discharge on the estimation, 7 days was chosen as a timeframe for the model given that our average LOS was 6.5 days, and the majority of infants (65% of cesarean, 75% of vaginal deliveries) had outcome data for ≥5 days. A 4-degrees of freedom B-spline was used for the effect of hours of life to generate nonlinear percentile curves.³⁷  Estimates for median percent weight loss at 24, 48, and 72 hours of life were obtained and compared to published nomograms of nonopioid exposed infants²⁹  by using 95% confidence intervals (CI), which were estimated by using the percentile method (2.5^th and 97.5^th percentiles) from 500 nonparametrically bootstrapped samples.⁴¹  Comparison of the median weight loss percentage across the hospital stay between high and low caloric supplementation infants was also performed by using the quantile regression approach, adjusting for infant’s ethnicity (Hispanic, yes or no), race (Black, White, Other), sex (male or female), BW, and gestational age at birth. Distributions of maximum percent weight loss and associated weight nadir, as well as LOS were compared among different feeding types and groups of caloric supplementation by using the Kruskal-Wallis test, followed by posthoc Dunn’s multiple comparisons test. Statistical hypothesis tests were conducted at the 2-sided α of .05. Quantile regressions for longitudinal data were implemented using lqmm package in R.^38,39  Descriptive analyses and between group comparisons were implemented using Prism GraphPad Version 8.4.1.

This study was approved by the Yale University institutional review board under protocol ID 2000021711.

Of 744 infants ≥35 weeks gestational age diagnosed with NOWS during the study period, 330 (41.6%) met inclusion criteria (Fig 1). Of included infants, 71.2% were delivered vaginally, 68.5% were born to mothers using methadone, and 20.3% to mothers on buprenorphine. Median LOS was 4.9 days; 94.6% did not receive pharmacologic treatment (Table 1). Characteristics and outcomes of included infants are presented in Table 1.

Figure 2A presents percentile weight loss curves for vaginally-delivered infants with NOWS. Median percentage weight loss for infants in our cohort at 24 hours was 4.3% (95% CI: 3.3–5.7) and at 48 hours was 6.9% (95% CI: 5.8–8.5) (Fig 2A), compared to 2.1% (P < .001) and 2.9% (P < .001), respectively, in nonopioid-exposed infants, according to published early weight loss nomograms.²⁹  Of vaginally-delivered infants in our cohort, 28.3% (95% CI: 22.7–34.5) reached >10% weight loss during hospitalization.

Figure 2B presents percentile curves for infants with NOWS delivered via cesarean delivery. Median percent weight loss of cesarean-delivered infants at 24 hours was 4.1 (95% CI: 2.0–6.5), at 48 hours was 6.5% (95% CI: 4.1–9.1), and at 72 hours was 7.2% (95% CI: 4.7–9.9) (Fig 2B), compared to 2.9% (P < .001), 3.7% (P < .001) and 3.5% (P < .001), respectively, in nonopioid-exposed infants, according to published early weight loss nomograms.²⁹  Likewise, 25.2% (95% CI: 16.9–35.2) of cesarean-delivered infants with NOWS lost >10% BW during hospitalization. Overall, 27.9% of infants with NOWS lost >10% BW.

Of newborns in our cohort, 3.0% were exclusively breastfed, 37.3% exclusively formula-fed, and 59.6% were combination-fed. Of all infants, 21.8% received the majority of their feeds as breast milk (including those exclusively breastfed) for the first 7 days of hospitalization. Maximum weight loss was lower in exclusively formula-fed infants compared to combination-fed infants who received a majority as breast milk (median 8.2% vs 9.4%, P = .01) (Fig 3A). Overall, median maximum weight loss for all infants was 8.6% (IQR 6.9–10.3).

The day of life of median weight nadir was not significantly different between the feeding groups and overall was 3.0 days (IQR 2.3–4.0) (Fig 3B). Median LOS for exclusively formula-fed infants or combination-fed infants who received a majority of feeds as formula (5.1 days and 5.0 days) was significantly longer than combination-fed infants who received a majority as breast milk (4.6 days, P = .006 and P = .009, respectively) as well as those who were exclusively breastfed (3.9 days, P = .02 and P = .03) (Fig 3C). Overall, the median LOS was 4.9 days (IQR 4.1–5.5).

Of infants in our cohort, 39.4% received increased caloric density feeds, with 15.4% of those starting increased caloric density feeds on day 0 or 1. Infants who lost significantly more weight were more likely to receive increased caloric density feeds (Fig 3D and Supplemental Fig 4).

Of infants who received increased caloric density feeds, those who received increased caloric density feeds earlier were found to have statistically significant differences in the 75 percentile for both maximal weight loss (5.5% vs 8.5%, P = .03) and LOS (6.2 days vs 7.2 days, P = .05) than those who received increased caloric density feeds later (Fig 3D and 3F). In addition, those who received feeds earlier were found to have earlier weight nadirs (5.3 days vs 5.5 days, P = .03) (Fig 3E).

Of infants in our cohort, 3.9% were fed via NGT for a median of 6.0 days (IQR 2.7–7.0) (Table 2). Of NGT-fed infants with NOWS, median maximum weight loss was 12.4% (IQR 11.0–13.9); median weight nadir was 5.5 (IQR 4.4–6.1); median LOS was 10.0 days (IQR 7.2–10.9); and 23.1% of infants with an NGT received morphine (Table 2).

Our study is the first to demonstrate weight loss patterns of infants with NOWS managed by using the ESC approach. These percentile curves can assist providers with expectations of weight loss in infants with NOWS in the newborn period when management decisions are based on the ESC approach, and they give a starting framework for weight loss in an infant going through withdrawal.

We show that compared to weight loss patterns of nonopioid-exposed infants,^28,29  infants with NOWS have substantially greater weight loss and continue to lose weight for longer periods of time. In our cohort, 27% of infants with NOWS lost >10% BW compared with only ∼0.1% of nonopioid-exposed infants, albeit the LOS for nonopioid exposed infants were shorter than those in our study.²⁹  The increased weight loss in the NOWS population still hospitalized past 5 days since birth is likely because of withdrawal signs such as hypertonicity and tremors that increase energy use and a tendency toward feeding difficulties that can compromise caloric intake. Consistent with nonopioid-exposed infants, vaginally-delivered infants with NOWS lost less weight than cesarean-delivered infants.

To date, few studies have examined weight loss patterns in infants with NOWS, and those that did used different management approaches, including the modified FNASS and modified Lipsitz scoring protocols, resulting in high percentages of infants receiving pharmacologic treatment (ranging from 45%–76%).^17,18,20  In contrast, only 5.4% of infants in our study were treated pharmacologically. The ESC approach, which encourages rooming-in and provides consistent delivery of nonpharmacologic interventions, allowed for analysis of weight loss patterns of a higher proportion of infants that were not treated with pharmacologic agents. Additionally, previous studies only evaluated infants exposed to methadone, whereas our study also included infants exposed to buprenorphine and short-acting opioids.

Compared to infants with NOWS managed with different approaches, our results suggest that infants managed with the ESC approach have similar maximum weight loss but reach their weight nadir more quickly and closer to that of nonopioid-exposed infants. Infants in our study had a median maximal weight loss that was within range of other studies of infants with NOWS (8.5% vs 4%–9.7%).^17,18,20  The median weight nadir overall for our cohort (3.0 days) is comparable to untreated infants in a previous study¹⁷  but earlier compared to ∼ 5 to 7 days in other published studies.^18,21,43  As infant weight loss is among the factors considered for newborn hospital discharge readiness, our results show that infants managed by the ESC approach at our institution begin to regain birth weight earlier than infants managed by other approaches, which can impact birth hospital LOS. Although our early weight loss percentile curves for infants delivered vaginally seems to depict infants continuing to lose weight at 7 days (168 hours of life), the majority (87%) were discharged by this point; thus, this downtrend in weight may only represent infants demonstrating more severe symptoms of NOWS. Future studies are warranted to assess the relationship of severity of NOWS symptoms and weight loss. Additionally, our findings underscore the importance of close discharge follow-up for this vulnerable population. All infants monitored for NOWS discharged from our institution are discharged from the hospital with pediatrician follow-up within 72 hours, a birth-to-3 referral, developmental clinic follow-up, and visiting nurse referral services. We have previously shown that infants managed using the ESC approach did not have adverse events, including transfer to the ICU from the inpatient unit, seizures, and readmissions within 30 days.^8,9 

We compared our weight loss percentile curves to previously published nomograms from nonopioid-exposed formula-fed infants because of the low rates of breastfeeding in our population. However, consistent with other studies of infants with NOWS,^20,28,29,44  we found that majority breast milk–fed infants experienced more weight loss but had a shorter LOS compared to those that were majority formula-fed. Barriers to breastfeeding in this population could include poor weight gain misinterpreted as poorly established lactation, inconsistent messaging from health care providers around medical and psychiatric comorbidities, and limited financial resources,^44–47  among others.

In our cohort, the small percentage of infants who received NGT feeds had a greater median weight loss than the overall median weight loss for infants with NOWS; despite this, only a small percentage received morphine. In general, infants losing a substantial amount of weight and showing signs of discoordinated suck and/or sleepiness received NGTs rather than morphine. Although NGT feeds have not been well described in the literature for infants with NOWS, NGTs are frequently used in premature infants in the NICU because of difficulty with coordinating sucking, swallowing, and breathing related to prematurity, which are signs also seen in infants with NOWS. Our data demonstrate that NGTs have been used at our center in infants with NOWS to assist with feeding discoordination.

Our findings show that those infants who received increased caloric density feeds also had significantly more weight loss than those receiving normal caloric density feeds (whether breast milk or formula). This result is likely because of selection bias, because most providers started increasing caloric density in infants who were losing more weight. However, infants who started on increased caloric density feeds earlier on day 0 or 1 (before significant weight loss being recorded) had lower maximal weight loss, earlier weight nadirs, and shorter lengths of stay than those who were started increased caloric density feeds on day 2 or later, although we were not specifically powered to find these effect sizes. Our results suggest further studies are warranted and necessary to elucidate effects of early increase in caloric density of feeds on weight loss in infants with NOWS managed by ESC.

Our study has several limitations. First, we did not have a standardized approach to feeding, particularly in regard to timing of introduction of high-calorie formulas. Infants who received morphine, NGT feeds, and increased caloric supplementation were included in the percentile curves. Because the study was retrospective and observational in design, we were unable to determine weight loss percentiles had the infants not received interventions of NGT feeds, morphine, or increased caloric supplementation. However, our supplemental figure separates those infants who received increased caloric density feeds and those who received normal caloric density feeds, with a similar slope of weight loss. We felt it was important because we wanted to show the overall trajectory of weight loss for all opioid-exposed infants rather than only those whose weight loss trajectory did not require any intervention. Second, we did not extend the study to the posthospitalization period; many of our infants were discharged before regaining BW, thus it is possible that our findings could be different had we obtained data for seven-day follow up on all infants. However, given that we modeled the weight percentiles of hospitalized infants with NOWS, the assumption of MAR for the primary outcome of weight loss during the hospital stay is strengthened given that we adjusted our models for other covariates and used all available outcome data for any given infant. Readers may wonder if the weight loss estimation at later time points on the graph (for example, at 7 days) is reflective of only infants at that time point. The reader can be assured that this would only be the case had we estimated the quantiles of the observed weight loss only and reported those data over time. However, using the statistical model with the MAR assumption for missingness, the estimation algorithm for our model in essence creates percentile curves that are informed by both observed and unobserved outcomes which are only dependent on the previously observed weights and covariates; thus, we believe our findings describe and are representative of the expected weight loss percentile curves for hospitalized infants at our center at each time point noted. Third, the weight loss percentile curves are derived from a single center, which limits its generalizability. Multicentered studies should be performed in the future to establish multicenter percentiles for weight loss in infants with NOWS. Lastly, a portion of the time period studied coincided with the second half of implementation of the ESC approach. Although most elements of the approach were in effect throughout the study period, “as needed” medication dosing was not added until 2015.

In this study, we show the first demonstration of weight loss patterns of infants hospitalized with NOWS using a management approach focused on nonpharmacologic interventions. Compared to weight loss patterns of nonopioid-exposed infants, opioid-exposed infants have substantially greater weight loss and continue to lose weight for longer periods of time, which can place them at risk for increased morbidity and health care use. We also demonstrate that compared with opioid-exposed infants managed by other approaches, infants managed by the ESC approach begin to regain birth weight earlier, which can impact birth hospital LOS. Further, we show that infants with NOWS who started on higher calorie feeds within the first 48 hours had lower maximal weight loss, earlier weight nadirs, and shorter lengths of stay compared to those who were started on higher calorie feeds later, suggesting a feasible, cost-effective approach to improving weight management. Definitive studies to address optimal feeding approaches starting at birth in infants with NOWS are necessary with a focus on both short-term outcomes and long-term neurodevelopmental outcomes.

The authors thank the resident and nursing staff of the inpatient unit at YNHCH and Caitlin Partridge, Yale New Haven Health Joint Data Analytics Team. The authors thank Julia Rosenberg, MD, and Elizabeth Dong Nguyen, MD, PhD, for their critical reading of the manuscript.