To compare head circumference (HC) in neonates treated for neonatal abstinence syndrome (NAS) with control neonates without antenatal opioid exposure.
Our prospective cohort study ran from April 1, 2014, through December 31, 2016. Newborns treated for NAS delivered from well-dated pregnancies ≥34 weeks’ gestation were compared with newborns who were nonopioid exposed and matched for race, parity, mode of delivery, and gestational age. All mothers underwent serial antenatal urine drug testing. A minimum of 754 study participants were needed (377 in each group) to demonstrate an increase in the proportion of newborns with HCs less than or equal to the 10th percentile from 10% in controls to a minimum of 20% in NAS newborns with 90% power.
A total of 858 neonates were enrolled (429 NAS cases and 429 controls). Mean HC for cases was 33.04 cm (±1.9 cm) compared with 33.99 cm (±2.0 cm) for controls (P < .0001). Among the 429 NAS cases, the mothers of 372 (87%) were on opioid medication-assisted treatment. For NAS cases, 30.1% (95% confidence interval: 25.8%–34.7%) had an HC less than or equal to the 10th percentile (129 of 429 neonates), and 8.2% (95% confidence interval: 5.8%–11.2%) had an HC less than or equal to the third percentile (35 of 429 neonates). Multivariate analysis was used and determined that only chronic opioid use during gestation resulting in a neonate who was NAS treated was a significant risk factor for the observed smaller HC.
Chronic opioid use during pregnancy sufficient to cause NAS was associated with smaller HCs at birth. Most mothers were on opioid agonist medication-assisted treatment, which is currently the recommended treatment option during pregnancy.
In a few published studies, authors have suggested that chronic opioid drug use during pregnancy is associated with smaller newborn head circumferences at birth. However, in these retrospective studies, researchers did not fully account for other potentially confounding antenatal drug exposures.
In this prospective study in a large series of well-dated pregnancies, we found that antenatal opioid exposure was associated with reduced head circumference at birth after controlling for other antenatal drug exposures as verified by maternal drug testing.
Opioid use disorder (OUD) in pregnancy has risen in the United States, resulting in an increase in neonatal abstinence syndrome (NAS).1 The current American College of Obstetricians and Gynecologists Committee Opinion recommendation for treating OUD in pregnancy is to manage the patient with medication-assisted treatment (MAT) by using either methadone or buprenorphine.2 The rationale for opioid MAT stems from evidence of maternal benefit (eg, decreased withdrawal symptoms, improved prenatal care, and reduced risk of recidivism), although neonates have a high risk of NAS, a treatable newborn condition.2 In a few studies, however, researchers have reported smaller newborn head circumferences (HCs) at delivery and smaller brain volumes in the chronic opioid–exposed newborn compared with controls,3,–7 which might adversely affect the newborn’s long-term neurodevelopmental outcome.
To our knowledge, there are limited prospective data on neonatal HC of newborns who develop NAS in which researchers have controlled for the type of maternal opioid drug as well as other antenatal drug exposures. Pregnant women may have coexposures to sedatives, stimulants, marijuana, gabapentin, alcohol, tobacco, selective serotonin reuptake inhibitors (SSRIs), and common antiemetics. In a previous retrospective study from our institution, researchers did find an association of smaller HCs in newborns treated for NAS compared with controls; but, because of the retrospective nature of the investigation, accurate information in which authors detailed the type of maternal opioid and other coexposures and provided better pregnancy dating were lacking.6
We report results of a prospective study we designed to determine if chronic opioid use during pregnancy, sufficient enough for the neonate to require treatment for signs of NAS, is associated with smaller HCs compared with newborns who are not exposed to opioids.
Methods
Our prospective cohort study included births from April 1, 2014, to December 31, 2016. The case and control populations came from the same clinic pregnancy population managed and delivered at the University of Tennessee Medical Center, Knoxville. Because NAS is a reportable condition in Tennessee, we standardized our case definition for a diagnosis of NAS using the Finnegan score.8 The criteria for starting pharmacologic treatment of NAS in our NICU consisted of any 2 consecutive Finnegan scores ≥10 or a single Finnegan score ≥12. The Institutional Review Board of the University of Tennessee Medical Center, Knoxville, approved creation of a database of all clinic pregnancies receiving their care and delivery at our institution that included delivery and newborn data. Each case was a neonate from the cohort, with a well-dated pregnancy that delivered at ≥34 weeks’ gestation, who was diagnosed with and treated for NAS in the NICU. Each control newborn was selected from the clinic pregnancy cohort that delivered shortly before or after the case newborn and was matched for race, parity, mode of delivery, and gestational age at delivery (±4 days). Control mothers did not have any illicit substance use on the basis of history and urine drug testing, had well-dated pregnancies, and delivered at ≥34 weeks’ gestation.
All pregnant women in the cohort had urine drug testing performed during prenatal care and at delivery. The mothers of the newborn NAS cases had a median of 6 tests, and mothers of controls had a median of 3 tests. Maternal data collection included demographics, medical and social history, and details about the delivery. We obtained information specific to the use of illicit drugs, tobacco, and alcohol as well as other antenatal medications, including gabapentin, SSRIs, and antiemetics, and confirmed reported drug and medication use by toxicology studies. Collected newborn data included sex, birth weight, and HC at birth.
Mothers of all cases and controls had several ultrasounds before delivery. Pregnancy dating for both cases and controls was based on the best obstetric estimate as recommended by the American College of Obstetricians and Gynecologists Committee Opinion.9 Only patients with an ultrasound before 20 weeks’ gestation were included. We used the American Academy of Pediatrics’ sex-specific nomograms of birth weight and HC by gestational age, which allowed categorization of infants with respect to the 10th and third percentiles as well as the determination of individual z scores.10 After delivery, all newborns were examined by a single group of physicians and pediatric nurse practitioners who employed a standardized approach for measuring HC. Specifically, if an infant demonstrated caput or molding, an examiner obtained the HC after these conditions had resolved. We excluded newborns with aneuploidy and major fetal anomalies.
The primary planned study end point was a comparison of neonatal HC between the cases and controls. Using data from the previous published retrospective study from the same institution,6 we determined that 754 study participants (377 in each group) were needed to detect a minimum increase in the percentage of neonates with an HC less than the 10th percentile from 10% in controls to 20% in infants with NAS by using an α (type I error) of .01 with 90% power. On the basis of the average number of NAS cases managed at our institution each month before study initiation, we anticipated that this investigation would span a minimum of 30 months. Given a study start date of April 1, 2014, we decided to enroll infants through December 31, 2016.
Statistical analysis involved simple percentages with Poisson binomial 95% confidence intervals (CIs). For comparisons, χ2 test, Fisher’s exact test, and Student’s t test were performed when applicable, with significance considered at P < .05. All comparisons were analyzed against a 2-sided alternative hypothesis. We planned to conduct multivariate analysis with the use of stepwise logistic regression if >1 significant factor were to be identified through univariate analysis. This would be performed with respect to determining which drug exposures occurred significantly more often in cases than controls and then with respect to determining which of these exposures might have been associated with the defined difference in HC. The study was reviewed and approved by the institutional review board at the University of Tennessee Medical Center, Knoxville.
Results
We enrolled and assessed 858 neonates (429 cases with NAS and 429 controls). With the data in Table 1, we show that mean HC in neonates with NAS was significantly less, by 9.5 mm, compared with controls. In addition, the proportions of NAS infants with HCs less than or equal to the 10th and third percentiles were significantly greater compared with control newborns. No other comparison was significantly different other than the length of newborn stay, which was anticipated because all of the cases by definition were treated in the NICU for NAS.
Demographics and Outcome Parameters of NAS Cases Compared With Controls (858 Total Neonates)
| Cases | Controls | P | |
|---|---|---|---|
| No. | 429 | 429 | — |
| Maternal age in y, mean (±SD) | 28.1 (±5.7) | 27.4 (±5.8) | .08 |
| White, n (%) | 410 (95.6) | 410 (95.6) | 1.0 |
| Multiparity, n (%) | 343 (80) | 343 (80) | 1.0 |
| Mode of delivery: vaginal, n (%) | 299 (69.7) | 299 (69.7) | 1.0 |
| Gestational age at delivery in wk, mean (±SD) | 38.3 (±1.5) | 38.2 (±1.6) | .35 |
| Gestational age range in wk | 340/7–424/7 | 340/7–420/7 | — |
| Gestational age <36 wk, n (%) | 39 (9.1) | 38 (8.9) | .90 |
| Male fetus, n (%) | 232 (54.1) | 217 (50.6) | .34 |
| Birth weight in g, mean (±SD) | 2938 (±478) | 3004 (±492) | .047 |
| Birth weight ≤10th percentile, n (%) | 61 (14.2) | 45 (10.5) | .12 |
| Birth weight ≤3rd percentile, n (%) | 18 (4.2) | 8 (1.9) | .07 |
| NICU length of stay in d, mean (±SD) | 21.3 (±5.7) | 3.1 (±1.0) | <.0001 |
| Neonatal HC in cm, mean (±SD) | 33.04 (±1.9) | 33.99 (±2.0) | <.0001 |
| Neonatal HC ≤10th percentile, n (%) | 129 (30.1) | 51 (11.9) | <.0001 |
| Neonatal HC ≤3rd percentile, n (%) | 35 (8.2) | 8 (1.9) | <.0001 |
| Chronic maternal medical disordersa, n (%) | 71 (16.6) | 93 (21.7) | .07 |
| Gestational or pregestational diabetes, n (%) | 33 (7.7) | 44 (10.3) | .23 |
| Cases | Controls | P | |
|---|---|---|---|
| No. | 429 | 429 | — |
| Maternal age in y, mean (±SD) | 28.1 (±5.7) | 27.4 (±5.8) | .08 |
| White, n (%) | 410 (95.6) | 410 (95.6) | 1.0 |
| Multiparity, n (%) | 343 (80) | 343 (80) | 1.0 |
| Mode of delivery: vaginal, n (%) | 299 (69.7) | 299 (69.7) | 1.0 |
| Gestational age at delivery in wk, mean (±SD) | 38.3 (±1.5) | 38.2 (±1.6) | .35 |
| Gestational age range in wk | 340/7–424/7 | 340/7–420/7 | — |
| Gestational age <36 wk, n (%) | 39 (9.1) | 38 (8.9) | .90 |
| Male fetus, n (%) | 232 (54.1) | 217 (50.6) | .34 |
| Birth weight in g, mean (±SD) | 2938 (±478) | 3004 (±492) | .047 |
| Birth weight ≤10th percentile, n (%) | 61 (14.2) | 45 (10.5) | .12 |
| Birth weight ≤3rd percentile, n (%) | 18 (4.2) | 8 (1.9) | .07 |
| NICU length of stay in d, mean (±SD) | 21.3 (±5.7) | 3.1 (±1.0) | <.0001 |
| Neonatal HC in cm, mean (±SD) | 33.04 (±1.9) | 33.99 (±2.0) | <.0001 |
| Neonatal HC ≤10th percentile, n (%) | 129 (30.1) | 51 (11.9) | <.0001 |
| Neonatal HC ≤3rd percentile, n (%) | 35 (8.2) | 8 (1.9) | <.0001 |
| Chronic maternal medical disordersa, n (%) | 71 (16.6) | 93 (21.7) | .07 |
| Gestational or pregestational diabetes, n (%) | 33 (7.7) | 44 (10.3) | .23 |
—, not applicable.
Includes hypertensive conditions, lupus, chronic renal disease, and others.
By study design, none of the controls were born to mothers with a history of antenatal exposure to opioids, sedatives, or stimulants, as validated by maternal urine drug testing. In Table 2, we show that through univariate analysis, we identified a significantly higher exposure to antenatal marijuana, gabapentin, tobacco, and SSRIs among the NAS cases. Cases and controls had similar exposure to promethazine and alcohol. Using NAS as the constant dependent variable, we performed a backward stepwise multivariate regression analysis to determine if any individual exposure was a significant independent risk factor for neonates treated for NAS. The model included opioid drug use in general, buprenorphine, methadone, benzodiazepine, stimulant, marijuana, gabapentin, SSRI, and tobacco usage. The analysis revealed that opioid drug exposure in general and buprenorphine, methadone, benzodiazepine, and marijuana exposure were significantly associated with NAS at P < .01. Tobacco, SSRI, stimulant, and gabapentin exposure were not significant independent risk factors.
Other Medication or Drug Use in NAS Cases Compared With Controls (858 Total Neonates)
| Cases | Controls | P | |
|---|---|---|---|
| No. | 429 | 429 | — |
| Opioid drug use, n (%) | 429 (100) | 0 (0) | <.0001 |
| Buprenorphine | 320 (74.6) | 0 (0) | <.0001 |
| Methadone | 52 (12.1) | 0 (0) | <.0001 |
| Benzodiazepine usage, n (%) | 90 (21.0) | 0 (0) | <.0001 |
| Stimulant usage, n (%) | 27 (6.3) | 0 (0) | <.0001 |
| Marijuana usage, n (%) | 199 (46.4) | 112 (26.1) | <.0001 |
| Gabapentin usage, n (%) | 61 (14.2) | 14 (3.3) | <.0001 |
| SSRI usage, n (%) | 54 (12.6) | 33 (7.7) | .02 |
| Promethazine usage, n (%) | 118 (27.5) | 97 (22.6) | .12 |
| Smoking, n (%) | 314 (73.2) | 221 (51.5) | <.0001 |
| Positive alcohol screen result, n (%) | 21 (4.9) | 19 (4.4) | .87 |
| Cases | Controls | P | |
|---|---|---|---|
| No. | 429 | 429 | — |
| Opioid drug use, n (%) | 429 (100) | 0 (0) | <.0001 |
| Buprenorphine | 320 (74.6) | 0 (0) | <.0001 |
| Methadone | 52 (12.1) | 0 (0) | <.0001 |
| Benzodiazepine usage, n (%) | 90 (21.0) | 0 (0) | <.0001 |
| Stimulant usage, n (%) | 27 (6.3) | 0 (0) | <.0001 |
| Marijuana usage, n (%) | 199 (46.4) | 112 (26.1) | <.0001 |
| Gabapentin usage, n (%) | 61 (14.2) | 14 (3.3) | <.0001 |
| SSRI usage, n (%) | 54 (12.6) | 33 (7.7) | .02 |
| Promethazine usage, n (%) | 118 (27.5) | 97 (22.6) | .12 |
| Smoking, n (%) | 314 (73.2) | 221 (51.5) | <.0001 |
| Positive alcohol screen result, n (%) | 21 (4.9) | 19 (4.4) | .87 |
By study design, the control group was negative for opioid drugs, benzodiazepine, and stimulant usage.
In examining the 429 NAS cases, 372 (87%) were on opioid agonist MAT (320 [75%] on buprenorphine and 52 [12%] on methadone), with the remaining 13% on other opioid drugs. There were 35 newborns with an HC less than or equal to the third percentile (8.2%; 95% CI: 5.8%–11.2%). There were 129 cases (30.1%; 95% CI: 25.8%–34.7%) with an HC less than the 10th percentile. Comparing these with the 300 NAS cases with a normal HC, we found no differences in exposure to buprenorphine, methadone, other opioid drugs, amphetamines, cocaine, benzodiazepines, marijuana, gabapentin, SSRIs, tobacco, or alcohol (Table 3).
Substance Use in the 429 NAS Cases Comparing the Small HC Group With the Normal HC Group (N = 429)
| HC <10th Percentile | Normal HC | P | |
|---|---|---|---|
| No. | 129 | 300 | — |
| Opioid agonist MAT, n (%) | 114 (88.4) | 258 (86.0) | .61 |
| Buprenorphine | 94/114 (82.5) | 226/258 (87.6) | .25 |
| Methadone | 20/114 (17.5) | 32/258 (12.4) | .25 |
| Opioid street drug usage, n (%) | 15 (11.6) | 42 (14.0) | .61 |
| Benzodiazepine usage, n (%) | 24 (18.6) | 66 (22.0) | .51 |
| Amphetamine usage, n (%) | 6 (4.7) | 11 (3.7) | .83 |
| Cocaine usage, n (%) | 3 (2.3) | 7 (2.3) | .99 |
| Marijuana usage, n (%) | 58 (45.0) | 141 (47.0) | .77 |
| Gabapentin usage, n (%) | 20 (15.5) | 41 (13.7) | .73 |
| SSRI usage, n (%) | 18 (14) | 36 (12.0) | .69 |
| Smoking, n (%) | 96 (74.4) | 218 (72.7) | .80 |
| Positive alcohol screen result, n (%) | 5 (3.9) | 16 (5.3) | .63 |
| HC <10th Percentile | Normal HC | P | |
|---|---|---|---|
| No. | 129 | 300 | — |
| Opioid agonist MAT, n (%) | 114 (88.4) | 258 (86.0) | .61 |
| Buprenorphine | 94/114 (82.5) | 226/258 (87.6) | .25 |
| Methadone | 20/114 (17.5) | 32/258 (12.4) | .25 |
| Opioid street drug usage, n (%) | 15 (11.6) | 42 (14.0) | .61 |
| Benzodiazepine usage, n (%) | 24 (18.6) | 66 (22.0) | .51 |
| Amphetamine usage, n (%) | 6 (4.7) | 11 (3.7) | .83 |
| Cocaine usage, n (%) | 3 (2.3) | 7 (2.3) | .99 |
| Marijuana usage, n (%) | 58 (45.0) | 141 (47.0) | .77 |
| Gabapentin usage, n (%) | 20 (15.5) | 41 (13.7) | .73 |
| SSRI usage, n (%) | 18 (14) | 36 (12.0) | .69 |
| Smoking, n (%) | 96 (74.4) | 218 (72.7) | .80 |
| Positive alcohol screen result, n (%) | 5 (3.9) | 16 (5.3) | .63 |
We then performed a forward stepwise logistic analysis of the primary outcome variable (HC ≤10th percentile) using each of the exposures that was found to be significantly different between the cases and controls in the backward multivariate analysis. The only exposure that was associated with a significant risk of an HC less than or equal to the 10th percentile was chronic antenatal opioid exposure resulting in a neonate treated for NAS. The relative risk for an HC less than or equal to the 10th percentile was 2.53, with an adjusted odds ratio of 3.19 (95% CI: 2.23–4.56; P < .0001), whereas the relative risk for an HC less than or equal to the third percentile was 4.38, with an adjusted OR of 4.67 (95% CI: 2.14–10.20; P < .0001).
In many studies, researchers analyze HC solely as a continuous variable. In this study, however, we prospectively made the primary outcome measure (for HC with respect to the 10th and third percentiles) a dichotomous value. However, to provide the fullest depiction of the differences in HC between the NAS and control newborns, we also determined the z score for the HC of each infant on the basis of gestational age at delivery and plotted the data for the cases and controls to create Gaussian curves (Fig 1). In the distribution of HC among NAS cases, a simple shift to the left was demonstrated, compared with controls. With this finding, the possibility is raised that antenatal exposure to opioids sufficient enough to result in NAS might have a deleterious effect of similar magnitude on HC in every NAS case, even if HC was in the normal range.
We designed our study to compare neonates who sustained chronic opioid exposure that resulted in NAS treatment with control newborns who were not exposed to opioids. In this prospective study, we did not include 216 neonates with antenatal exposure to opioids who did not develop NAS. These latter newborns were born to mothers on full-maintenance opioid MAT, to mothers on other opioid drugs, and to mothers who were tapering their opioid drug dosage but remained on a reduced dose of opioid at the time of delivery. The demographics and the gestational age at delivery (38.1 ± 1.6 weeks) did not differ between these infants and the NAS cases in this study. The rates of benzodiazepine, gabapentin, marijuana, and tobacco usage were also similar to the NAS cases. The mean HC was 33.27 ± 2.0 cm, the rate of an HC less than or equal to the 10th percentile was 26.9%, and the rate of an HC less than or equal to the third percentile was 6.9%. For the subset of infants without NAS whose mothers were on MAT, 71 were on buprenorphine, and 42 were on methadone.
Discussion
We found that chronic opioid use during pregnancy sufficient enough to result in NAS was associated with a significant decrease in HC at birth. Newborns with NAS had more frequent coexposures to benzodiazepines, stimulants, marijuana, gabapentin, tobacco, and SSRIs than did non-NAS controls. However, when we assessed these individual drug exposure relationships within the NAS newborn population alone, we found that none of the coexposures was a significant risk factor for a smaller HC at birth. It is noteworthy that the majority of NAS cases were born to mothers on opioid agonist MAT, which is currently the recommended treatment option during pregnancy.
To our knowledge, this is the first large prospective study in which researchers compare HC in newborns treated for NAS with well-matched controls in which all mothers had well-dated pregnancies and in which determinations of exposure were made by using both history and multiple urine drug tests. Our findings were similar to the retrospective material previously reported from our institution that had stimulated the current, more-rigorous, and prospective investigation.6 In that previous study, 332 neonates treated for NAS were compared with controls, and greater proportions of infants with NAS had a newborn HC less than or equal to the 10th and less than or equal to the third percentiles. In that study, authors were unable to make a definitive conclusion about the possible effect of antenatal opioid exposure on neonatal HC because data pertaining to other possible confounding drugs and exposures were incomplete. Arlettaz et al,3 in a retrospective study of 86 newborns exposed to methadone, reported a 13% incidence of HC less than the third percentile, but 62% were also exposed to additional drugs, primarily cocaine. Bier et al11 reported significantly smaller HC z scores for neonates exposed to a high dose of methadone (>100 mg daily dose) compared with newborns exposed to a low dose of methadone or buprenorphine. Additionally, in a few small studies, researchers have reported smaller brain volumes and abnormal white matter development in older children (ages 9–14) who received treatment of NAS after birth compared with controls, suggesting that children may not experience “catch-up” growth in HC through early adolescence.4,5,7
Many intrauterine perturbations that affect fetal growth elicit a variety of compensatory mechanisms that allow the fetus to preserve the development and growth of the central nervous system. To the contrary, our data revealed a significant 2.8% reduction in mean HC and a smaller proportional reduction of 2.2% in mean birth weight. Because newborn HC is an indirect measure of brain volume, further research is necessary to determine if this finding increases the risk for long-term neurodevelopmental delay. Current data on the long-term neurodevelopmental effects of antenatal opioid exposure are conflicting. Authors of recent studies involving brain imaging have suggested that there is abnormal neural tract development and myelination in newborns assessed at 2 to 6 weeks of age as well as at 12 to 15 years of age.12,13 In a 2008 publication, researchers reported an increased risk for neurodevelopment delay through early childhood in infants who were exposed to chronic maternal opioid use, whereas in a meta-analysis from 2014, researchers reported no significant impairment with respect to neurobehavioral outcomes in exposed children.14,15 However, in the past 20 years, there are numerous reports of impaired psychomotor development, visual-motor disturbances, increased attention-deficit/hyperactivity disorder, lower IQ scores, lower Griffith MD scores, lower executive function, weaknesses in language and cognition, and poorer high school performance in children who were treated for NAS at birth.16,–24
Few data are available to evaluate whether detoxification during pregnancy of patients with OUD lowers the risk for a reduced HC and what effect detoxification might have on long-term childhood outcome. Haabrekke et al25 reported a mean HC of 33.9 (±1.9) cm in 78 newborns who were substance exposed, a mean HC of 34.8 (±1.5) cm in 22 newborns delivered to mothers who underwent full detoxification during pregnancy (resulting in no NAS), and a mean HC of 35.4 (±1.2) cm in control newborns who were not exposed. The difference in mean HC was significant between the drug-exposed and control groups but not between the detoxified and control groups.25 Walhovd et al26 reported that the MRI brain volumes did not differ at a mean age of 4.5 years in 12 newborns delivered to mothers who detoxified during pregnancy, resulting in an absence of NAS, compared with 12 controls. At present, detoxification during pregnancy is not recommended in the primary management of OUD in pregnancy2 and has only been reported in a few locations (in both inpatient and outpatient settings) in the United States and other countries.27,–31 If the practice of detoxification during pregnancy expands, research regarding newborn head size and long-term outcomes on these infants and children is of paramount importance.
Not all neonates chronically exposed to opioid drugs during pregnancy develop NAS postdelivery.32 The reason for this is not certain but may involve innate genetic differences, differences in placental metabolism, or other unknown factors. We did not include newborns who were born to women with OUD but who did not require treatment of NAS in this study. Our reason for not including these newborns in the current study was that we wished to ensure a minimum threshold level of antenatal opioid exposure and set that threshold as the development of NAS symptoms severe enough as to require treatment. Those neonates exposed to opioid drugs during pregnancy that do not develop NAS that requires treatment may represent a different population in which the nature of opioid exposure and its effect are more speculative. Because our study has now revealed that neonates treated for NAS have a significantly smaller mean HC compared with controls, further prospective research to assess the impact of opioid exposure in newborns who do not develop NAS is still needed. Such newborns may be born to women on maintenance MAT, women on illicit opioids, or women who underwent partial or complete opioid detoxification. Such variable antenatal exposure may provide valuable information about whether there is a critical predelivery time period of exposure that is associated with a reduction in neonatal HC or whether the opioid effect is cumulative over time.
Any investigation in which researchers seek to define a relationship between a drug and an outcome is limited to the extent to which the exact amount and timing of drug exposure are known. Although the women in our practice did have urine drug testing performed, we could not precisely quantify the opioid dosage in all cases or know if other drugs not included in the urine drug test played a role or were consumed at a time outside the performance limitations of the test. However, 179 substances were analyzed in the urine drug test used in this study.33 Although we could not quantitate a dose-response relationship for each mother-infant dyad, we do provide the best evidence to date that chronic antenatal opioid exposure sufficient enough to lead to NAS that requires treatment is associated with a significant reduction in neonatal HC independent of other coexposures.
Conclusions
Chronic opioid use during pregnancy that results in a newborn with signs of NAS that requires treatment appears to reduce head growth in utero compared with nonopioid-exposed controls, as assessed by comparisons of mean HC, proportions of HC less than or equal to the 10th or third percentiles, or distribution of z scores of HC by gestational age. The majority of NAS cases were born to mothers receiving opioid agonist MAT, which is currently the recommended treatment option during pregnancy. Analyses of non-NAS newborns exposed to maintenance MAT, illicit opioids, or lesser doses and/or durations of MAT because of partial or complete maternal detoxification are needed to determine if the current recommended management of OUD in pregnancy should be revisited.
Dr Towers conceptualized and codesigned the study, coordinated and supervised the maternal and neonatal data collection, conducted the initial analyses, and drafted the initial manuscript; Dr Hyatt supervised the maternal and neonatal data collection and critically reviewed and revised the manuscript; Ms Chernicky and Ms Chattin conducted the maternal and neonatal data collection and critically reviewed the manuscript; Drs Visconti and Fortner codesigned the study, supervised the maternal data collection, and critically reviewed and revised the manuscript; and all authors approved the final manuscript as submitted.
FUNDING: Partially funded through the Blue Cross Blue Shield Research Foundation.
COMPANION PAPER: A companion to this article can be found online at www.pediatrics.org/cgi/doi/10.1542/peds.2018-3376.
Dear Editor,
We would like to thank Desai et al, Lipshaw et al, and Jones et al, for raising important questions about our study. Regarding Desai, et al, they question whether selection bias among our cases might have accounted for our findings. They question the possibility of greater exposures among cases to poverty, inadequate prenatal care, or higher maternal stress.
Poverty has been associated with an increased risk of low birth weight and preterm birth. However, we have not found evidence to suggest that poverty level or reduced access to prenatal care have an adverse impact on neonatal head circumference as a function of gestational age. Nevertheless, on further review of our prospectively collected data, there were no significant differences between cases and controls in the percentage of mothers insured by Medicaid (91.6% v. 90.4%, P=0.63) or the number of prenatal visits (11.1±2.9 v. 10.8±2.6, P=0.11). Given these univariate findings, it is not surprising that these variables did not achieve statistical significance when we repeated our multivariate analysis.
Regarding the question of maternal stress, it is estimated that over 500,000 pregnant women in the US per year have psychiatric illnesses.1 In our study, although treatment with selective serotonin reuptake inhibitors (SSRIs) was more common among cases than controls (Table 2: 12.6% v.7.7%), SSRIs were not an independent predictive factor in multivariate analysis. Stress may lead to poorer overall pregnancy outcomes, but we are not aware that stress alone results in smaller head circumferences.2
A compelling finding of our study (that matched cases and controls for gestational age) was that opioid exposure had a greater adverse effect on head circumference than birth weight as reflected by a larger percentage reduction in the mean measurements (2.8% v. 2.2%) and higher percentages with measurements below the 10th (30.1% v. 11.9%) and 3rd percentiles (8.2% v. 1.9%). If exposures to poverty or maternal stress had been greater among cases, it is likely these factors would have had a proportionately greater adverse effect on birth weight than on head circumference, due to compensatory fetal mechanisms that preserve head growth in the face of challenges to somatic growth.
Our study reports an association between opioid exposure and reduced head circumference only in a subset of infants known to have experienced significant cumulative exposure because we documented chronic maternal opioid use and because the infants developed signs of NAS severe enough to result in treatment. The 3 sets of commentator’s express skepticism that all infants with chronic antenatal opioid exposure will demonstrate smaller head circumferences; we agree. This was neither our conclusion nor our implication. Shortly after we found an apparent association, we initiated the next logical step in the discovery process and are nearing conclusion of a similar prospective study on pregnancy outcomes of women with chronic opioid exposure whose neonates did not require treatment for NAS. The most secure way to assess for an independent effect of fetal opioid exposure would be through a prospective randomized trial; however, such a trial is not easily feasible because pregnant women can and should exercise their right to make an informed choice about their management.
Lipshaw et al, raises an additional question of whether we may have reversed causation and suggest instead that smaller head circumferences may predispose opioid-exposed infants to more severe signs of NAS. While there might be biological credibility to this hypothesis, the degree of leftward shift of head circumference z-scores and the preservation of the Gaussian distribution of this variable in cases compared with controls, does not provide strong evidence for this alternative hypothesis.
Jones et al, have four questions regarding our study. First, they state that we posited a backward causal chain: that neonatal NAS treatment leads to reduced head circumference. We did not. We selected cases to be newborns treated for NAS because it is reasonable to assume that these neonates sustained chronic and significant fetal opioid exposure. In other words, we used a clinical marker as the criterion for a threshold of exposure. To be clear, we cannot know what the fetal exposure truly was in any newborn without signs of NAS who is born to a mother on chronic opioid treatment. Total maternal opioid dosage consumed is difficult to verify with certainty and clinicians do not measure maternal serum opioid levels. More importantly, fetal opioid levels are unknowable. As Jones notes in her publication, “Metabolism of methadone by the placenta, fetus, and infant are likely to play a role in the variable expression of NAS among opioid-exposed infants and may additionally be affected by other factors such as genetics.”3
In this first criticism, Jones et al also raise a concern that neonatal treatment for NAS might have been driven by newborn head circumference. This is categorically untrue and presupposes that clinicians might have unethically chosen to treat babies with smaller head circumferences preferentially to “drive” the outcome of a hypothesis. As the study reports, the diagnosis of NAS was based strictly on Finnegan scores and head circumference was not part of this algorithm. As a practical matter, 70% of cases had head circumferences greater than the 10th percentile, such that a clinician would not flag these babies to have a subnormal head circumference.
We do not fully understand the thrust of Jones et al’s second criticism. We did address in great detail the differences in head circumference between cases and controls. Regarding the third question, we agree that the maternal level of opioid agonist medication at delivery does not correlate well with NAS. For instance, a low level may reflect lower maternal dosing or high maternal dosing with significant placental metabolism (producing lower fetal exposure); or higher maternal dosing with the last dose remote from delivery or pregnancies with decreased placental metabolism (producing higher fetal exposure). In the first instance, the risk of NAS is low; in the second instance, the risk is certainly greater. Yet, one cannot logically use this observation to conclude, as implied, that the presence of NAS does not correlate with a threshold level of fetal opioid exposure.
Fourth, Jones et al, question whether poorer nutrition among cases might have affected head growth. Because this was a prospective study and included only those patients with good dates (ultrasounds prior to 20 weeks of gestation), we were able to collect good data on maternal weight gain that showed no difference between cases and controls (19.1+8.9 v. 20.3+11.5 pounds, P = 0.09). No patients in the study lost weight. Weight gain was not significant in repeating our multivariate analysis. We must point out again that poor nutrition is associated with greater adverse effects on somatic growth compared to head growth, in contradistinction to our findings.
Finally, we fully concur that all patients using illicit opioid drugs should be placed into methadone or buprenorphine medication assisted treatment (MAT) programs; this is the mainstay of our own program. However, pregnant women who are stable in a MAT program should be educated about other management options and be provided intense support, comprehensive follow-up, and in the context of any research study, have maternal, fetal, and neonatal outcomes prospectively studied. We were very clear not to extrapolate our findings to conclude that any opioid exposure reduces neonatal head circumference. We are in agreement that social determinants of health are key influences on neonatal and childhood outcomes. However, we believe that it is unconscionable, given the association that we have described, to deny the importance of additional well-designed studies to test the hypothesis that significant chronic opioid exposure impacts fetal head growth.
The authors report no conflicts of interest.
References:
1. Use of psychiatric medications during pregnancy and lactation. ACOG Practice Bulletin No. 92. American college of Obstetricians and Gynecologists Obstet Gynecol 2008;111:1001-20.
2. Shaw JG, Asch SM, Kimerling R, et al. Posttraumatic stress disorder and risk of spontaneous preterm birth. Obstet Gynecol 2014;124:1111-9.
3. Jones HE, Jansson, LM, O'Grady KE, Kaltenbach, K. The relationship between maternal methadone dose at delivery and neonatal outcome: methodological and design considerations. Neurotoxicol Teratol. 2013;39:110–5.
Dear Editor,
Towers et al. report that in a prospective cohort study of infants with neonatal abstinence syndrome (NAS), infants with NAS had smaller head circumferences than control infants (1). Both the authors and the accompanying editorial by Hudak and Makker conclude that chronic opioid use during pregnancy is associated with smaller head circumferences (2). We are concerned that these conclusions are overreaching given the authors selection of cases and controls. One of the strengths of a prospective cohort study is the ability to accurately assess exposure and to reduce risk of selection bias. Rather than using prenatal exposure to opioids as the case definition, the authors use elevated Finnegan scores. Although commonly used for the diagnosis of NAS, Finnegan scores have never been validated and are poor markers for the "severity" of in-utero exposure (3-5). By introducing the confounding and potential selection bias associated with a post-natal case definition, the authors may have reversed their causality, i.e., infants with smaller heads are more likely to demonstrate severe withdrawal symptoms. If the authors wished to make conclusions about the effects of prenatal opioid exposure, they should compare infants exposed to unexposed infants. If the authors wish to make conclusions about the association of head circumference to NAS, they should compare opioid exposed infants with severe withdrawal symptoms to asymptomatic opioid exposed infants. Towers et al. do neither.
1. Towers CV, Hyatt BW, Visconti KC, Chernicky L, Chattin K, Fortner KB. Neonatal Head Circumference in Newborns With Neonatal Abstinence Syndrome. Pediatrics. 2019;143(1).
2. Hudak ML, Makkar K. Fetal Opioid Exposure and Smaller Birth Head Circumference: Cause for Concern? Pediatrics. 2019;143(1).
3. Cleary BJ, Donnelly J, Strawbridge J, et al. Methadone dose and neonatal abstinence syndrome-systematic review and meta-analysis. Addiction. 2010;105(12):2071-84.
4. Jansson LM, DiPietro JA, Elko A, Velez M. Maternal vagal tone change in response to methadone is associated with neonatal abstinence syndrome severity in exposed neonates. J Matern Fetal Neonatal Med. 2007;20:677–685.
5. McCarthy JJ, Leamon MH, Stenson G, Biles LA. Outcomes of neonates conceived on methadone maintenance therapy. J Subst Abuse Treat. 2007 Dec 11;
Towers and colleagues asserted that “Chronic opioid use during pregnancy that results in a newborn with signs of NAS that requires treatment appears to reduce head growth in utero compared with non-opioid-exposed controls”.(1) However, multiple design/methodological issues seriously undercut this conclusion.
First, they posit a backward causal chain - neonatal NAS treatment leads to reduced head circumference. Given a backwards causal chain, a strong relationship between levels of maternal opioid use and NAS treatment is necessary. Also concerning is that NAS treatment occurs after measuring neonatal head circumference. It is unknown if head circumference partly drove NAS treatment.
Second, control women had no illicit substance use history. However, opioid use disorder (OUD) in pregnant women commonly occurs in the context of multiple social determinants of health not frequently operative in women without illicit substance use. Comparing head circumference in neonates treated- versus untreated-for-NAS may be necessary to reach a tentative conclusion regarding the relationship of fetal environment to incidence of treatment for NAS but is not sufficient to conclude that maternal OUD causes reduced head circumference. Further, the magnitude of the difference in head circumference between groups was not addressed.
Third, maternal opioid use was not measured. Use-nonuse is inferred from NAS, a ‘stand-in variable’. Since, research shows that NAS is unrelated to maternal opioid agonist medication levels at delivery (2), NAS is a poor stand-in variable for maternal opioid use, as it assesses neither quantity nor timing of opioids and NAS may be due to licit or illicit use of other substances (e.g., benzodiazepines, SSRIs, tobacco etc.).(3)
Fourth, ‘crude classification of the observations’ lowers power and leads to biased tests of significance.(4) Beyond the crude classification problem, there is also a threat to the validity of statistical conclusions by omitted variables. Notably, inadequate maternal nutrition, a known causative factor of smaller head size,(5) is not a covariate. Given that higher methadone dose has been shown to be related to larger head circumference some attempt to use a morphine-equivalent dose in the subsample taking methadone or buprenorphine seems warranted.
Summary
Overreaching and misleading conclusions are harmful to mothers and children. Understanding problems in fetal development needs to focus on identifying the maternal, social, and environmental causal factors that produce developmental problems such as reduced head circumference in the context of a model that acknowledges the complexity and interplay of causative factors. The use of NAS treatment to indicate maternal substance use causes harm to the child and family through labeling and misses the opportunity to investigate social determinants of health which contribute to fetal and later child development. Ignoring the complexity of influences on fetal development, and then attributing infant birth parameters to a single factor, fails to advance our understanding of the mother-child dyad, and detracts from optimal clinical care and future research.
References
1. Towers CV, Hyatt BW, Visconti KC, Chernicky L, Chattin K, Fortner KB. Neonatal Head Circumference in Newborns With Neonatal Abstinence Syndrome. Pediatrics. 2019;143(1) doi: 10.1542/peds.2018-0541
2. Jones HE, Jansson, LM, O'Grady KE, Kaltenbach, K. The relationship between maternal methadone dose at delivery and neonatal outcome: methodological and design considerations. Neurotoxicol Teratol. 2013;39:110–5. doi: 10.1016/j.ntt.2013.05.003
3. Kocherlakota P. Neonatal Abstinence Syndrome. Pediatrics. 2014;134(2):e547-61. doi: 10.1542/peds.2013-3524
4. Olsson U, Drasgow F, Dorans NJ. On The Robustness of Factor Analysis Against Crude Classification of the Observations. Multivariate Behavioral Research. 1979;14(4): 485-500.
5. Thame M, Wilks RJ, McFarlane-Anderson N, Bennett FI, Forrester TE. Relationship between maternal nutritional status and infant's weight and body proportions at birth. European Journal of Clinical Nutrition. 1997;51:134-138.
Dear Editor,
Towers and colleagues asserted that “Chronic opioid use during pregnancy that results in a newborn with signs of NAS that requires treatment appears to reduce head growth in utero compared with non-opioid-exposed controls”.(1) However, multiple design/methodological issues seriously undercut this conclusion.
First, they posit a backward causal chain - neonatal NAS treatment leads to reduced head circumference. Given a backwards causal chain, a strong relationship between levels of maternal opioid use and NAS treatment is necessary. Also concerning is that NAS treatment occurs after measuring neonatal head circumference. It is unknown if head circumference partly drove NAS treatment.
Second, control women had no illicit substance use history. However, opioid use disorder (OUD) in pregnant women commonly occurs in the context of multiple social determinants of health not frequently operative in women without illicit substance use. Comparing head circumference in neonates treated- versus untreated-for-NAS may be necessary to reach a tentative conclusion regarding the relationship of fetal environment to incidence of treatment for NAS but is not sufficient to conclude that maternal OUD causes reduced head circumference. Further, the magnitude of the difference in head circumference between groups was not addressed.
Third, maternal opioid use was not measured. Use-nonuse is inferred from NAS, a ‘stand-in variable’. Since, research shows that NAS is unrelated to maternal opioid agonist medication levels at delivery (e.g., 2-3), NAS is a poor stand-in variable for maternal opioid use, as it assesses neither quantity nor timing of opioids and NAS may be due to licit or illicit use of other substances (e.g., benzodiazepines, SSRIs, tobacco etc.). (4)
Fourth, ‘crude classification of the observations’ lowers power and leads to biased tests of significance.(5-6) Beyond the crude classification problem, there is also a threat to the validity of statistical conclusions by omitted variables. Notably, inadequate maternal nutrition, a known causative factor of smaller head size,(7) is not a covariate. Given that higher methadone dose has been shown to be related to larger head circumference(8) some attempt to use a morphine-equivalent dose in the subsample taking methadone or buprenorphine seems warranted.
Summary
Overreaching and misleading conclusions are harmful to mothers and children. Understanding problems in fetal development needs to focus on identifying the maternal, social, and environmental causal factors that produce developmental problems such as reduced head circumference in the context of a model that acknowledges the complexity and interplay of causative factors. The use of NAS treatment to indicate maternal substance use causes harm to the child and family through labeling and misses the opportunity to investigate social determinants of health which contribute to fetal and later child development. Ignoring the complexity of influences on fetal development, and then attributing infant birth parameters to a single factor, fails to advance our understanding of the mother-child dyad, and detracts from optimal clinical care and future research.
References
1. Towers CV, Hyatt BW, Visconti KC, Chernicky L, Chattin K, Fortner KB. Neonatal Head Circumference in Newborns With Neonatal Abstinence Syndrome. Pediatrics. 2019;143(1) doi: 10.1542/peds.2018-0541
2. Jones HE, Dengler E, Garrison A, et al. Neonatal outcomes and their relationship to maternal buprenorphine dose during pregnancy. Drug Alcohol Depend. 2013;134:414-417. doi: 10.1016/j.drugalcdep.2013
3. Jones HE, Jansson, LM, O'Grady KE, Kaltenbach, K. The relationship between maternal methadone dose at delivery and neonatal outcome: methodological and design considerations. Neurotoxicol Teratol. 2013;39:110–5. doi: 10.1016/j.ntt.2013.05.003
4. Kocherlakota P. Neonatal Abstinence Syndrome. Pediatrics. 2014;134(2):e547-61. doi: 10.1542/peds.2013-3524
5. Olsson U, Drasgow F, Dorans NJ. On The Robustness of Factor Analysis Against Crude Classification of the Observations. Multivariate Behavioral Research. 1979;14(4): 485-500.
6. Olsson U. Measuring Correlation in Ordered Two-Way Contingency Tables. Journal of Marketing Research. 1980;17(30):391-394.
7. Thame M, Wilks RJ, McFarlane-Anderson N, Bennett FI, Forrester TE. Relationship between maternal nutritional status and infant's weight and body proportions at birth. European Journal of Clinical Nutrition. 1997;51:134-138.
8. Doberczak TM, Thornton JC, Bernstein J, Kandall, SR. Impact of Maternal Drug Dependency on Birth Weight and Head Circumference of Offspring. Am J Dis Child. 1987;141(11):1163-1167.
Dear Editor,
We read the manuscript evaluating neonatal head circumference in newborns with neonatal abstinence syndrome (NAS) by Towers et al.,(1) with great interest. Prospective data collection that allowed the authors to accurately capture NAS and to incorporate toxicology results to ascertain medication exposure in pregnancy is a strength of this study. However, we are concerned that findings presented in this study are limited by selection bias introduced by a peculiar design choice. The authors explicitly excluded non-NAS cases from consideration for selection if an infant was exposed to any opioid (including opioid agonist therapy [OAT]), benzodiazepines, or stimulants. This restriction induced a perfect association between the exposure of interest (chronic opioid use) and NAS (seen in Table 2, 100% of the NAS cases vs 0% of the non-cases are exposed to opioids). Since the association between NAS and small head circumference can be potentially explained by unmeasured common causes, such as poverty, inadequate prenatal care, or maternal stress, the association observed between chronic opioid exposure and head circumference remains subject to bias.(2)
We suspect that the association between chronic opioid use and small head circumference observed by Towers et al. is likely to have resulted from conditioning on NAS and is non-causal. In other words, neonates with small head circumference were more likely to have had NAS than not, which could be due to unmeasured differences in sociodemographic factors between NAS cases and non-cases. Since non-NAS cases had zero probability of exposure to chronic opioid therapy by design, it is unsurprising that the authors found a positive association between chronic opioid therapy and small head circumference.
Next, the authors conclude that “chronic opioid use during pregnancy sufficient to cause NAS was associated with smaller HCs at birth;” however, the methodology employed is not sufficient for causal inference due to the selection bias described above. The conclusion then challenges the current standard of care of opioid agonist therapy for pregnant women. Opioid agonist therapies have been demonstrated to improve outcomes in women with OUD (e.g., reduction of overdose death)(3) and their infants (e.g., reducing preterm birth)(4) and they remain a vital tool in our armamentarium in the fight against the opioid epidemic. This study was not designed to determine if opioids cause smaller head circumference; therefore, we would urge readers to use caution when interpreting the conclusion from the study by Towers et al., and the implication that OAT may not be an appropriate treatment option for women with opioid use disorder
References:
1. Towers CV, Hyatt BW, Visconti KC, Chernicky L, Chattin K, Fortner KB. Neonatal Head Circumference in Newborns With Neonatal Abstinence Syndrome. Pediatrics. 2019;143(1).
2. Hernán MA, Hernández-Díaz S, Robins JM. A structural approach to selection bias. Epidemiology (Cambridge, Mass). 2004:615-625.
3. Sordo L, Barrio G, Bravo MJ, et al. Mortality risk during and after opioid substitution treatment: systematic review and meta-analysis of cohort studies. Bmj. 2017;357:j1550.
4. Fullerton CA, Kim M, Thomas CP, et al. Medication-assisted treatment with methadone: assessing the evidence. Psychiatr Serv. 2014;65(2):146-157.