Among 5053 mother-infant dyads affected by OUD, > 1 in 8 had documented prenatal fentanyl exposure. Though some contextual characteristics were similar between those with and without prenatal fentanyl exposure, those with documented exposure were more likely to be publicly insured, have their first prenatal care appointment in the third trimester, and have documented prenatal exposures to cannabis, sedatives, stimulants, nicotine and/or other nonopioid substances. Documented prenatal fentanyl testing and exposure increased over 20fold from 2014 to 2021. Overall, ~1 in 12 infants in the cohort had at least 1 of the 7 included infant features, most commonly microcephaly, male genital anomalies, and foot positioning deformities. Of these infants, the vast majority had only one feature documented, in contrast to the constellation of features observed among the 10 infants in the Wadman et al⁸ case series.

Prenatal fentanyl exposure was documented more commonly among infants with at least one of the included features compared with infants with none of these features. However, several unmeasured contextual factors limited our ability to causally attribute these features to fentanyl. First is the unknown drug purity. Adulterants are common in illicit drugs, including illegalized fentanyl,^5,18 though they are not systematically tested for or knowingly consumed.¹⁹ The potential role of unknown adulterants among infants with any of these features of interest remains unknown. Second, the need to consider potential effects of other prenatal substances and medication exposures is paramount when considering the association between fentanyl exposure and these infant features, which are hypothesized to be related to aberration in cholesterol metabolism.^8,10 It is unknown whether other opioid types can have similar effects on cholesterol biosynthesis.¹⁰ Antipsychotic medications may interrupt fetal cholesterol metabolism,^8,20 and OUD and mental health disorders frequently co-occur,²¹ so future analyses should include these medications. In this analysis, we also observed a higher prevalence of other prenatal substance exposures among dyads with, versus without, prenatal fentanyl exposure. Given the known association between substances like alcohol²² and nicotine²³ and birth defects, and fentanyl’s increasing role as an adulterant in some nonopioid illicit substances,²⁴ future studies may consider the role of other substance exposures. Third, other contextual factors²⁵ including environmental exposures, disproportionately affect people affected by OUD and have been linked to several features in this study, including cleft palate and limb defects.^26,27 Taken together, the fundamental question of whether fentanyl is driving the observed infant features in Wadman et al (2023),⁸ or if those infant features may be the result of other exposures, separately or in combination with fentanyl, remains unanswered.

The more than 20-fold increase in fentanyl exposure over time in our analysis of dyads affected by OUD emphasizes the importance of further understanding its impacts. From 2014–2021, fentanyl use geographically expanded in the United States,²⁴ potentially exacerbated by the COVID-19 pandemic,²⁸ but data-related considerations also likely influenced our observed increase. Awareness of fentanyl in the illicit drug supply increased during this timeframe, in part due to harm reduction efforts,²⁹ meaning mothers affected by OUD later in the surveillance period may have been more likely to know they were using fentanyl or a drug adulterated by fentanyl, report it to their provider, and have medical record documentation. This time period also saw significant increases in the availability of laboratory testing for fentanyl in clinical settings,³⁰ potentially increasing the number of pregnancies with fentanyl exposure identified across time. However, MAT-LINK clinical sites qualitatively reported heterogeneous fentanyl testing protocols, even in recent years, where some clinical sites conducted routine urine drug tests, including fentanyl tests, among all pregnant women with a substance use disorder, whereas others did not conduct in-house fentanyl toxicology tests and decreased toxicology testing over time. Therefore, prenatal fentanyl exposure might be underestimated at some clinical sites. In addition, before October 2020, no ICD-10-CM code to document fentanyl exposure was available; therefore, some instances of prenatal fentanyl exposure might be missed by our exposure definition, as health care providers were limited to coding different or broader synthetic opioid classes.³¹ Manually abstracted indicators of prenatal fentanyl use are additionally subject to underreporting, as manual abstractors relied on information documented in the EHR to determine whether prenatal fentanyl exposure occurred for each infant. Thus, MAT-LINK data likely contains many misclassified³² and unreported instances of prenatal fentanyl exposure. However, our ability to leverage abstracted medical information in addition to laboratory testing facilitated the identification of 667 infants with documented prenatal fentanyl exposure, 224 of whom were identified exclusively through abstracted medical information.

Some infant features in this analysis might be more prevalent than general US estimates; however, direct comparison is difficult due to different ascertainment methodologies, feature definitions, and random error from smaller counts. We observed 0.5% of infants having any cleft palate with or without cleft lip, whereas the National Birth Defect Prevention Network (NBDPN) estimated that from 2016 to 2020, 0.06% of live births had cleft palate only and 0.1% had cleft lip with or without cleft palate.³³ Our observed prevalence of talipes equinovarus was also 3 times NBDPN’s estimate of 0.2%.³³ When we defined microcephaly using a standardized head circumference ≤ 3%, we identified almost three times the number of infants that we would expect. Among both infants with and without documented prenatal fentanyl exposure, the median birth weight was lower than average. Further research is needed to understand whether the prevalence of these features is elevated among infants with prenatal opioid, specifically fentanyl, exposure, while considering the additional potential impacts of maternal health, prenatal care, prescription medications, polysubstance use, and environmental and other exposures. Further research is also needed to understand why feeding difficulties were more common among infants with prenatal fentanyl exposure, compared with those without, in this analytic sample. It is possible that severity of NOWS presentation could explain at least part of this relationship, and warrants further consideration using data that can capture severity of NOWS and feeding difficulties with a standardized approach.³⁴

Attributing fentanyl as a cause for the observed features has real-world consequences for mothers with substance use disorders, so it is important that evidence justifying any attribution is strong. These implications include stigma from health care providers,³⁵ inadequate prenatal care access,³⁵ and other punitive consequences,³⁶ particularly among under-resourced communities.³⁷ While efforts continue to investigate potential causal mechanisms of the observed features in the Wadman et al⁸ case series, providers can take steps to support dyads exposed to fentanyl and other opioids. MOUD is the standard of care for treating OUD during pregnancy, including those who use fentanyl.³⁸ In our data, we observed MOUD was documented more frequently during pregnancy among dyads with versus without prenatal fentanyl exposure. This might be because receiving MOUD might result in more health care system interactions, leading to more opportunities for documentation of fentanyl exposure. Though fentanyl use often warrants additional MOUD prescribing considerations due to risk of buprenorphine-precipitated withdrawal from fentanyl, particularly during pregnancy,³⁹ MOUD is still associated with reductions in adverse pregnancy and infant features compared with no or shorter treatment and is an important prenatal care provision for dyads affected by OUD.³⁸

This analysis is subject to several limitations. Because we leveraged an existing data source, rather than standardized physical examinations of each infant, not all features documented in the Wadman et al⁸ case series, including cholesterol metabolism, facial features, adducted thumbs, and single palmar creases, could be examined in MAT-LINK data. Infant features that are included are subject to misclassification. For example, micrognathia maps imperfectly to ICD codes, and rocker bottom feet do not have an ICD-9-CM specific code. Thus, we expect that these infant features will be documented less reliably than others, such as cleft palate, in the EHR. In addition, some included features might not have been recognized by a physician or documented using ICD codes. Several limitations in the ascertainment of prenatal fentanyl exposure using EHRs are also described above. While we could not verify that all fentanyl exposures ascertained were nonpharmaceutical, including epidural medications,⁴⁰ we did not include prescribed fentanyl in our exposure definition and relied on sources of information likely to indicate nonpharmaceutical use. This applies to other types of prenatal substance exposures captured in our analysis as well, such as cannabis, stimulants, and benzodiazepines. Due to limited fetal information on stillborn infants in MAT-LINK, we restricted our analysis to live-born infants. Our results might underestimate the impact of these infant features, particularly among those with prenatal fentanyl exposure, if these features contribute to stillbirth, especially in the presence of prenatal fentanyl exposure. Our results are not generalizable to the entire US population of maternal-infant dyads affected by OUD, as MAT-LINK only encompasses 7 clinical sites. These data do not fully represent the spectrum of maternal characteristics, substance use patterns, and genetic profiles of all women accessing perinatal OUD services in the United States. As a secondary analysis of sentinel surveillance data designed to broadly capture dyad features, this report does not equip us to definitively substantiate a syndrome attributed to prenatal fentanyl exposure. Strengths of our analysis include a large sample of infants among whom fentanyl exposure was relatively common. This helped us contextualize the extent to which signals of infant features reported in the Wadman et al⁸ case series are observed in a population of susceptible infants across 7 clinical sites, including cases where only some features of interest are expressed, even if we are not yet equipped with this data source alone to substantiate the plausibility of a fetal fentanyl syndrome.

The paucity of data about fentanyl’s fetal effects is concerning amid an ongoing fentanyl use crisis. The case series published by Wadman et al⁸ proposed a fetal syndrome associated with prenatal fentanyl exposure, accompanied by a plausible biological mechanism^8,10; however, considering alternative or interaction mechanisms is prudent given the contextual differences we observed between those with and without documented prenatal fentanyl exposure. Several features of interest to this proposed syndrome were present but rare in our sample, and < 1% of infants had at least 2 of these features documented. Further investigation incorporating multilevel factors is needed to examine all potential fac tors that may be driving the features observed in Wadman et al⁸ before substantiating a fetal fentanyl syndrome.

Supplemental Digital Content is available for this article. Direct URL citations are provided in the HTML and PDF versions of this article on the journal’s website, www.journaladdictionmedicine.com.