Buprenorphine and Methadone Discontinuation during Pregnancy and the Postpartum Period: A Nationwide Cohort study

Abstract

Background:

Treatment of pregnant patients with opioid use disorder (OUD) with methadone or buprenorphine is crucial for maternal and neonatal safety. While several clinical trials have demonstrated higher treatment discontinuation for buprenorphine compared with methadone outside of pregnancy, evidence during pregnancy and the postpartum period is limited.

Objective:

To compare treatment discontinuation between buprenorphine and methadone during pregnancy and with follow-up extended through 1-year postpartum.

Methods:

Using nationwide Medicaid data, we conducted a cohort study of pregnant patients who initiated methadone or transmucosal buprenorphine (±naloxone) for OUD during the first trimester. The primary outcome was treatment discontinuation, defined as a ≥60-day treatment gap, with alternative definitions for discontinuation explored in sensitivity analyses. Hazard ratios (HRs) were estimated using Cox proportional hazards regression with propensity score overlap weighting to control for confounding. Subgroup analyses were conducted stratified by buprenorphine-alone versus the buprenorphine/naloxone combination product, each compared to methadone.

Results:

We identified 696 pregnant patients who initiated methadone and 1538 who initiated buprenorphine in the first trimester. Compared to methadone initiators, buprenorphine initiators were more likely to discontinue treatment during pregnancy (32.8% for buprenorphine vs. 25.6% for methadone; weighted HR 1.41, 95%CI 1.15–1.72) and through 1-year postpartum (58.8% vs. 49.0%; HR 1.37, 95%CI 1.19–1.57). For patients initiating the buprenorphine/naloxone combination, the HR was 1.73 (95%CI 1.36–2.19) during pregnancy and 1.56 (95%CI 1.30–1.86) through 1-year postpartum. For patients initiating buprenorphine-alone, the HRs were 1.14 (95%CI 0.90–1.46) and 1.23 (95%CI 1.04–1.46) respectively. Varying the treatment gap used to define discontinuation in sensitivity analyses yielded consistent results.

Conclusion:

Pregnant patients initiating transmucosal buprenorphine during early pregnancy are more likely to discontinue treatment than those initiating methadone, but treatment discontinuation is high on both treatments. Our findings highlight the importance of identifying and addressing barriers to treatment retention among pregnant patients with opioid use disorder.

Introduction

In the United States, the prevalence of opioid use disorder (OUD) among pregnant individuals has markedly increased, affecting almost one percent of all pregnancies.^1–3 Treatment with medications for opioid use disorder (MOUD) is crucial and strongly recommended for pregnant patients to prevent opioid withdrawal symptoms disruptive to maternal and fetal well-being, to reduce illicit opioid use and other substance use during pregnancy, to facilitate the continuation of prenatal care, and to improve birth outcomes.^4,5 First-line MOUD treatments include methadone, a full agonist at mu-opioid receptors, and buprenorphine, a partial agonist at mu-opioid receptors.

Retention in MOUD treatment is associated with substantial reductions in both all-cause and overdose-related mortality.⁶ Evidence outside of pregnancy consistently shows that buprenorphine is less effective than methadone in retaining patients in treatment.^7,8 However, there are limited comparative data regarding treatment retention during pregnancy and the postpartum period for buprenorphine versus methadone.⁹ In the Maternal Opioid Treatment: Human Experimental Research (MOTHER) trial, the largest trial of pregnant patients with opioid dependence to date (N=175 participants), the proportion of patients who discontinued treatment was greater in the buprenorphine group than in the methadone group.¹⁰ Yet, there is a dearth of evidence in the real world setting,¹¹ where the administration of these medications differs significantly. In the United States, methadone treatment of OUD is typically provided to patients under daily supervision via federally-regulated opioid treatment programs (OTP), while buprenorphine is most commonly prescribed through routine outpatient practice.¹² In addition, although the postpartum period is a critical time with a greater risk of return to use and opioid overdose than during pregnancy,^4,13 to date, only one study has compared retention between methadone and buprenorphine in the postpartum period based on data from a single state.¹⁴ Comparative data on secondary outcomes associated with treatment retention, such as adequacy of prenatal care and adverse OUD-related events (e.g., overdose), are also lacking.

The aim of this study was to compare discontinuation of MOUD treatment between buprenorphine and methadone during pregnancy and with follow-up extended through the first year postpartum in a nationally representative publicly-insured population, and to explore implications with respect to adequacy of prenatal care and OUD- and mental health-related adverse events.

Methods

Data source and study cohort

We conducted a cohort study using nationwide US Medicaid data, which include publicly insured beneficiaries from 47 states and Washington, D.C. The majority of pregnant patients with OUD in the US receive insurance coverage through Medicaid.^1,2 The pregnancy cohort was constructed using data from 2003 to 2018 (the most current data at the time of study conduct), with the cohort construction process detailed in prior publications.¹⁵ The Medicaid data comprise patient demographics, insurance enrollment information, and medical claims data, including diagnoses and procedures from hospital, outpatient, and emergency visits, as well as prescriptions for outpatient medications.

The source cohort included pregnancies ending in live birth with continuous enrollment from at least 3 months before the date of the estimated last menstrual period (LMP) to 1 month after delivery. The primary study cohort included pregnant patients who initiated MOUD treatment with methadone or buprenorphine (±naloxone) between LMP and LMP+90 days (eFigure 1). Treatment initiation was defined as no evidence of treatment during the 3 months prior to LMP. We focused on treatment initiation, as opposed to any use, given the well-known time trends in treatment adherence. In the absence of complete information on the duration of prior MOUD use, mixing initiators and prevalent users may distort the results by under-representing patients who discontinued early in their treatment course, as only prevalent users who continue beyond the early period would be included in the study (known as depletion of susceptibles).¹⁶

Exposure

Methadone treatment was identified using Healthcare Common Procedure Coding System (HCPCS) codes, while buprenorphine treatment was primarily identified through prescription fill records supplemented by HCPCS codes (eTable 1).^17,18 The treatment initiation date was assigned as the first prescription fill date or HCPCS code date between LMP and LMP+90 days. For prescriptions, the treatment record was defined based on the days’ supply (e.g., 30 days for a prescription with a 30-day supply). For HCPCS codes, the treatment record was defined as the date the code was recorded, assuming daily billing (i.e., each treatment record was 1 day). However, variations in billing patterns for methadone, such as daily, weekly, and monthly billing patterns, have been observed both in the literature¹⁹ and in our data. Consequently, grace periods (treatment gaps) were incorporated when defining treatment discontinuation to accommodate these variations in billing practices.

Treatment discontinuation

Time until treatment discontinuation was assessed (i) from MOUD initiation until end of pregnancy, and (ii) from MOUD initiation until 1-year post-partum or end of enrollment. In the main analysis, a 60-day gap was allowed between treatment records, and treatment was considered discontinued 60 days after the last treatment record (i.e., end of days’ supply or day of last administration code recorded) (eFigure 2).^14,20 Switching between methadone and buprenorphine (and between alternative formulations of buprenorphine) was allowed, as the clinical goal is to retain patients on any MOUD treatment. Specifically, for methadone initiators, treatment was considered continued if they stopped methadone but started buprenorphine within 60 days, and vice versa for buprenorphine initiators.

As an alternative approach, we estimated the proportion of months covered (PMC), defined as months with evidence of treatment (i.e., a HCPCS code was recorded or days of prescription supply overlapped with a given month) divided by the total months in the study period. Lack of treatment retention was defined as a PMC<80%. This approach captures the differences between patients who discontinued treatment and never restarted versus those who temporarily discontinued but later resumed treatment.

Adequacy of prenatal care

Adequacy of prenatal care was defined in two ways. First, prenatal care was considered adequate if a second trimester anatomical ultrasound was performed during the recommended time window: gestational weeks 18 to 20 (strict definition) or gestational weeks 17 to 22 (lenient definition) (eTable 2).²¹ Second, we calculated a modified version of the Adequacy of Prenatal Care Utilization (APNCU) index,^22,23 which compares the number of actual prenatal care visits with the recommended visits, accounting for the timing of MOUD initiation (eMethods). Prenatal care was defined as inadequate if initiated after the fourth gestational month or if fewer than 50% of the recommended visits were received; prenatal care was defined as intermediate if 50–79% of the recommended visits were received.

OUD- and mental health-related adverse events

Four OUD- and mental health-related adverse events were assessed, including opioid overdose, any drug overdose, self-harm, and mental health-related hospitalization. Definitions, diagnosis codes and their accuracy measures from the literature are summarized in eTable 2.^24–27 Follow-up started the day after MOUD initiation and ended with the first occurrence of the specific adverse event being studied, death, or end of follow-up (defined as [i] end of pregnancy, or [ii] 1-year postpartum or end of enrollment).

Covariates

A comprehensive range of patient-level confounders was considered, including demographics, timing of treatment initiation, markers of OUD severity, medical conditions associated with OUD, substance use, mental health conditions, chronic comorbid conditions, other medication use, and healthcare utilization. For analyses evaluating OUD- and mental health-related adverse events, prior histories of such events before MOUD initiation were also considered. The covariates and their assessment periods are summarized in eTable 3. In addition, a sensitivity analysis was performed, restricted to pregnant patients with complete socioeconomic status information, to adjust for county-level socioeconomic status proxies (i.e., proximity to metropolitan area, education level, poverty rates, and unemployment rates assessed through linking zip codes with data provided by the US Department of Agriculture).²⁸

Statistical analyses

Patient characteristics were summarized as means with standard deviations for continuous variables and as counts with percentages for categorical variables. Standardized mean differences were used to evaluate covariate imbalances. An absolute standardized mean difference >0.1 was considered evidence of imbalance.²⁹ To control for confounding, a logistic regression model incorporating all covariates was used to derive the propensity score and each individual was weighted by the probability of receiving the opposite treatment to what they actually received.^30,31 This overlap weighting approach mitigates the influence of individuals with extreme weights and ensures perfect balance of covariates.^30,31

For treatment discontinuation and OUD- and mental health-related adverse events, survival curves and cumulative incidences were estimated using the Kaplan–Meier method, and crude and weighted hazard ratios (HR) were estimated using Cox proportional hazards regression. For lack of treatment retention defined as PMC<80% and adequacy of prenatal care, crude and weighted odds ratios (OR) were estimated using logistic regression. We calculated 95% confidence intervals (CI) using robust variance to account for the weighting. To simplify the presentation, we use the more general term “risk” to describe the hazard ratio and odds ratio results.

Sensitivity and subgroup analyses

For treatment discontinuation, several sensitivity analyses were performed to test the robustness of the results: (i) shortening the treatment gap from 60 days to 32 days (a 32-day gap was used instead of a 30-day gap to accommodate the monthly billing patterns of methadone, specifically for those with HCPCS codes occurring on the same day each month), (ii) extending the treatment gap from 60 days to 90 days, (iii) censoring at treatment switch, (iv) censoring at initiation+9 months instead of at delivery for analyses restricted to the pregnancy-period to account for differential length of follow-up for preterm versus term deliveries, (v) restricting to patients with complete follow-up (i.e., without preterm deliveries and without administrative censoring throughout the entire postpartum year), (vi) restricting to patients without hospitalization during follow-up to account for incomplete information on inpatient medication use, and (vii) restricting to patients from states with Medicaid coverage for both methadone and buprenorphine³².

Subgroup analyses were conducted stratified by initiation of buprenorphine-alone versus the buprenorphine/naloxone combination product, each compared to methadone. Switching between buprenorphine-alone and buprenorphine/naloxone was not considered a treatment discontinuation.

In addition to the primary study cohort consisting of pregnant patients who initiated methadone or buprenorphine between LMP and LMP+90 days, two additional study cohorts were defined. The second study cohort included pregnant patients who initiated methadone or buprenorphine between LMP and LMP+240 days or delivery, whichever occurred first. While the primary study cohort was restricted to initiators early in pregnancy to ensure sufficient follow-up for evaluation of treatment discontinuation during pregnancy, the larger size of the second study cohort provided increased precision to evaluate rare outcomes, such as OUD- and mental health-related adverse events. Furthermore, a third study cohort was created to specifically examine treatment discontinuation during the postpartum period, regardless of treatment during pregnancy. This cohort included pregnant patients who received methadone or buprenorphine treatment within 60 days after delivery and were followed from the date of the first prescription or administration code after delivery up to 1-year postpartum or until the end of enrollment. Differences in the analyses and covariates between the primary, second, and third study cohorts are described in eFigure 1, eTable 3, and eTable 4.

All analyses were conducted using SAS Software, version 9.4 (SAS Institute). This study was approved by the institutional review board of Brigham and Women’s Hospital, which granted a waiver of informed consent.

Results

Cohort characteristics

Among 2,522,450 live birth deliveries, 696 pregnant patients who initiated methadone and 1538 who initiated buprenorphine (±naloxone) between LMP and LMP+90 days were included in the primary cohort (eFigure 1). Compared to methadone initiators, buprenorphine initiators were more likely to be White, to have a history of opioid-related hospitalizations, and to have other substance use disorders and mental health disorders, as well as prescriptions for other medications. Buprenorphine initiators were more likely to live in rural areas and in regions with lower education levels, higher poverty rates, and higher unemployment rates (Table 1 and eTable 5). After applying overlap weighting, the two treatment groups achieved perfect balance on all covariates by design.

Table 1.

Selected Characteristics of Pregnant Patients Initiating Buprenorphine versus Methadone in the First Trimester (Primary Cohort)

Characteristics	Buprenorphine		Methadone		Standardized Difference†
	N	%	N	%
No. of initiators	1538		696
Age, mean (SD)	27.7	4.7	28.2	4.9	−0.11
Race
Black	50	3.3	55	7.9	−0.20
Hispanic	64	4.2	54	7.8	−0.15
Other/unknown	91	5.9	51	7.3	−0.06
White	1333	86.7	536	77.0	0.25
Region
Midwest	641	41.7	192	27.6	0.30
Northeast	517	33.6	193	27.7	0.13
South	243	15.8	191	27.4	−0.29
West	137	8.9	120	17.2	−0.25
Multiple gestation	24	1.6	14	2.0	−0.03
Markers of OUD history and severity
Opioid related ED visit	212	13.8	93	13.4	0.01
Opioid related hospitalization	267	17.4	73	10.5	0.20
Substance use
Alcohol use disorder	140	9.1	41	5.9	0.12
Other substance use disorder	352	22.9	93	13.4	0.25
Tobacco use	544	35.4	169	24.3	0.24
Residential substance use treatment	447	29.1	129	18.5	0.25
Mental disorders
ADHD	77	5.0	22	3.2	0.09
Anxiety	488	31.7	176	25.3	0.14
Bipolar disorder	183	11.9	74	10.6	0.04
Depression	415	27.0	147	21.1	0.14
PTSD	100	6.5	47	6.8	−0.01
Other medication use
Antibiotics	810	52.7	329	47.3	0.11
Antidepressants	563	36.6	196	28.2	0.18
Antiemetics	630	41.0	221	31.8	0.19
Antihypertensives	213	13.8	54	7.8	0.20
Anxiolytics	311	20.2	106	15.2	0.13
Benzodiazepines	255	16.6	132	19.0	−0.06
Other hypnotics	257	16.7	81	11.6	0.15
ADHD medications	103	6.7	27	3.9	0.13
Healthcare utilization
No. of outpatient visits, mean (SD)	5.6	8.0	7.8	13.8	−0.19
No. of ED visits, mean (SD)	0.8	1.7	0.9	1.8	−0.01
No. of hospitalizations, mean (SD)	0.1	0.4	0.1	0.5	0.02
County-level proxies for socioeconomic status (SES)
No. of initiators with complete SES information	1523		689
Proximity to metropolitan area
Metropolitan	1209	79.4	657	95.4	−0.50
Non-metropolitan or rural	314	20.6	32	4.6	0.50
Percentage of Population with ≤ High School Diploma *
Low	366	24.0	251	36.4	−0.27
Low to moderate	392	25.7	186	27.0	−0.03
Moderate to high	295	19.4	139	20.2	−0.02
High	470	30.9	113	16.4	0.35
Poverty Rate *
Low	454	29.8	254	36.9	−0.15
Low to moderate	384	25.2	191	27.7	−0.06
Moderate to high	402	26.4	132	19.2	0.17
High	283	18.6	112	16.3	0.06
Unemployment Rate *
Low	491	32.2	232	33.7	−0.03
Low to moderate	350	23.0	187	27.1	−0.10
Moderate to high	346	22.7	152	22.1	0.02
High	336	22.1	118	17.1	0.12

^*Categories based on quartiles.

^†All variables were perfectly balanced after overlap weighting (eTable 5).

Abbreviations: ADHD, attention-deficit/hyperactivity disorder; ED, emergency department; OUD, opioid use disorder; PTSD, post-traumatic stress disorder

A total of 1832 pregnant patients initiating methadone and 3596 initiating buprenorphine between LMP and LMP+240 days were included in the second cohort. Additionally, 2003 pregnant patients who received methadone treatment and 3714 who received buprenorphine treatment between delivery and delivery+60 days were included in the third cohort (eFigure 1). Differences in baseline characteristics between the two treatment groups in the second and third cohorts are presented in eTable 6 and 7.

Treatment discontinuation

The cumulative risk of discontinuation at the end of pregnancy in the propensity score weighted population was 32.8% for buprenorphine initiators, compared to 25.6% for methadone initiators. At one year postpartum, the cumulative risk of discontinuation increased to 58.8% for buprenorphine initiators and 49.0% for methadone initiators (Figure 1; eFigure 3 for crude results).

Figure 1. Weighted Kaplan-Meier Curves Comparing Treatment Discontinuation Between Buprenorphine (BUP) and Methadone (MET) in the Primary Study Cohort.

Note: The colored areas represent 95% confidence bands, calculated with the Hall-Wellner method and extending to the last event (treatment discontinuations) in each group.

Comparing buprenorphine to methadone, the corresponding weighted HR was 1.41 (95% CI 1.15–1.72) during pregnancy, and 1.37 (95% CI 1.19–1.57) through the first year postpartum (Figure 2; eTable 8 for crude HR). Results were consistent when varying the treatment gap and when changing the censoring criteria or restricting the study population as shown in Figure 2. Similarly, buprenorphine initiators were more likely to have a PMC<80% than methadone initiators: the weighted OR was 1.46 (95% CI 1.16–1.83) during pregnancy and 1.58 (95% CI 1.29–1.94) through 1-year postpartum.

Figure 2. Weighted Hazard Ratios (HRs) Comparing Treatment Discontinuation Between Buprenorphine (BUP) and Methadone (MET) in the Primary Study Cohort.

Abbreviations: PMC, proportion of months covered; SES, socioeconomic status

Patients who initiated buprenorphine/naloxone tended to begin treatment earlier in pregnancy than patients who initiated buprenorphine-alone (e.g., 56.4% of buprenorphine/naloxone patients started in the first month of pregnancy versus 6.3% of buprenorphine-alone patients) (eTable 9). When comparing the buprenorphine/naloxone combination vs. methadone, the weighted HRs for treatment discontinuation were 1.73 (95% CI 1.36–2.19) during pregnancy and 1.56 (95% CI 1.30–1.86) through 1-year postpartum. When comparing buprenorphine-alone vs. methadone, the HRs were 1.14 (95% CI 0.90–1.46) and 1.23 (95% CI 1.04–1.46) respectively (Figure 1, Figure 2, and eTable 8).

The results for the second, larger cohort were consistent with those from the primary cohort (eTable 10). Results from the third cohort, which focused on the postpartum period only, showed a slightly higher risk of treatment discontinuation for buprenorphine compared to methadone (42.3% vs. 36.9%; weighted HR = 1.18, 95% CI 1.07–1.31). There was no difference in treatment discontinuation between buprenorphine/naloxone and methadone (37.9% vs. 37.1%; weighted HR = 1.02, 95% CI 0.89–1.16), but an increased risk of discontinuation for buprenorphine-alone compared to methadone (44.9% vs. 35.9%; weighted HR = 1.32, 95% CI 1.18–1.48) (eFigure 4).

Adequacy of prenatal care and OUD- and mental health-related adverse events

Depending on the definition of inadequate prenatal care, buprenorphine initiators were either somewhat less likely than methadone initiators to receive inadequate prenatal care or there was no difference in risk (Figure 3, eTable 11).

Figure 3. Weighted Odds Ratios (ORs) Comparing Adequacy of Prenatal Care Between Buprenorphine (BUP) and Methadone (MET) in the Primary Study Cohort.

Abbreviations: APNCU, Adequacy of Prenatal Care Utilization

The number of OUD- and mental health-related adverse events was too small to draw meaningful conclusions when follow-up ended at delivery. When extending follow-up through the postpartum period, the estimates from the primary cohort remained imprecise, but results from the second cohort did not suggest a large difference in the risk of OUD- and mental health-related adverse events. (Figure 4, eTable 12). Similar results were observed when an extended look-back period was used to assess the history of OUD- and mental health-related adverse events (eFigure 5, eTable 13).

Figure 4. Weighted Hazard Ratios (HRs) Comparing OUD- and Mental Health-Related Adverse Events Between Buprenorphine (BUP) and Methadone (MET) in the Primary and Second Study Cohort.

Note: Cell sizes smaller than 11 for observed outcomes are suppressed in accordance with the Centers for Medicare and Medicaid Services’ cell size suppression policy.

Discussion

In this nationwide study, we evaluated treatment discontinuation and associated outcomes among patients initiating MOUD treatment during pregnancy. The main and sensitivity analyses consistently showed that patients who initiated buprenorphine (±naloxone) treatment in the first trimester were more likely to discontinue treatment during pregnancy and through one year postpartum compared to patients who initiated methadone. Despite the higher treatment discontinuation, buprenorphine initiators did not experience more inadequate prenatal care or more OUD- and mental health-related adverse events than methadone initiators, although estimates were imprecise. In addition, the analysis specifically focused on the postpartum period revealed that patients who received buprenorphine (±naloxone) treatment within 60 days after delivery were more likely to discontinue during the postpartum period than those who received methadone treatment.

Cumulative evidence from studies outside of pregnancy indicates a lower discontinuation rate for methadone compared to buprenorphine. For example, at one year, methadone showed lower discontinuation in both randomized control trials (pooled relative risk 0.82, 95% CI 0.68–0.98) and observational studies (pooled relative risk 0.73, 95% CI 0.63–0.85).⁷ Similarly, in the MOTHER trial which was conducted among pregnant patients (mean gestational age at treatment initiation: 16–19 weeks), the proportion of patients who discontinued treatment was higher in the buprenorphine group (33%) than in the methadone group (18%) at the end of pregnancy.¹⁰ Our study, which extends the existing evidence by evaluating pregnant patients who initiated treatment during the first trimester in the observational setting, also found that buprenorphine initiators had greater treatment discontinuation than methadone initiators during pregnancy and through the postpartum period. The difference in treatment discontinuation during pregnancy was less pronounced in our study compared to the MOTHER trial, potentially due to different medication administration methods in randomized control trials (double-blinded, with both medications given daily under observation in the study clinic) versus clinical practice (methadone was provided via federally-regulated opioid treatment programs; buprenorphine was prescribed through routine outpatient practice).¹⁰

The higher risk of discontinuation observed with buprenorphine versus methadone during pregnancy in our study was primarily driven by patients who initiated the buprenorphine/naloxone combination. Outside of pregnancy, the buprenorphine/naloxone combination is preferred because of its lower potential for misuse.^12,33 In contrast, during the 2003–2018 study period, buprenorphine-alone was recommended for OUD treatment during pregnancy due to limited safety data for the prenatal use of naloxone.^5,34 At that time, the available evidence on fetal safety was most extensive for methadone, followed by buprenorphine-alone, and least for buprenorphine/naloxone. These differences may influence prescribers’ treatment choices and the counseling provided to patients, potentially affecting patients’ motivation to continue or discontinue treatment during pregnancy. Our data also showed that patients who initiated buprenorphine/naloxone tended to begin treatment earlier in pregnancy than patients who initiated buprenorphine-alone. Therefore, it is possible that patients on buprenorphine-alone, who were more likely to begin their treatment after confirming their pregnancy, may have gained an understanding of the importance of MOUD during pregnancy based on physician consultation, and made an informed decision to start treatment during pregnancy, suggesting a higher likelihood of continuing treatment during pregnancy. It is also possible that, compared to buprenorphine/naloxone, buprenorphine-alone was more likely to be prescribed by obstetricians, who are also involved in providing prenatal care and may thereby enhance treatment retention. Conversely, patients on buprenorphine/naloxone may have initiated treatment before knowing they were pregnant. Upon discovering their pregnancy, out of concerns of fetal safety, the prescriber or patient may have (i) chosen to discontinue any MOUD treatment or (ii) attempted to switch to buprenorphine-alone, but a suboptimal switching process (e.g., due to unavailability or unaffordability) may have contributed to the complete discontinuation of all MOUD treatment.

The risk of return to use for patients with OUD is elevated in the postpartum period, which is potentially attributable to stress and sleep deprivation from newborn care, hormonal shifts, the possibility of postpartum depression, and change in focus from maternal to infant care.^4,35 This has prompted guidelines to emphasize the need for continued OUD treatment after childbirth.⁴ An earlier study using data from Massachusetts including patients who received MOUD in the delivery month or the month before found no difference in treatment discontinuation between buprenorphine (±naloxone) and methadone at one year postpartum.¹⁴ Results from our third study cohort (i.e., patients who received methadone or buprenorphine treatment within 60 days after delivery) showed a slightly higher risk of discontinuation in the postpartum period for buprenorphine (±naloxone) compared with methadone. The increased risk of discontinuation was observed in the buprenorphine-alone group, but not in the buprenorphine/naloxone group. Similar to the pregnancy period, the intention to switch treatment may play a role in treatment discontinuation during the postpartum period. Patients may attempt to switch from buprenorphine-alone to the buprenorphine/naloxone combination product after delivery, as guidelines typically recommend the combination product for non-pregnant patients and some state Medicaid programs limit buprenorphine-alone use to pregnant patient only,^12,18,36 but the switching process may have been suboptimal. It is also possible that buprenorphine-alone was more likely to be prescribed by obstetricians. Patients may face challenges finding another prescriber to continue their treatment as they transfer out of the care of the obstetrician. Alternatively, given the higher abuse liability of buprenorphine-alone compared to the buprenorphine/naloxone combination product,³³ patients on buprenorphine-alone may have been more likely to return to use and discontinue MOUD treatment during the postpartum period, when the direct harmful effects of substance use on the fetus are no longer a concern.

Our findings related to prenatal care adequacy align with prior studies, which generally also found no difference in the number of prenatal visits between methadone and buprenorphine, with some suggesting slightly higher adequacy of prenatal care in the buprenorphine group compared to the methadone group.^10,37,38

Several limitations should be considered when interpreting the results of our study. First, treatment records in claims database may not fully capture the actual use of MOUD. Misclassification is possible among patients treated with buprenorphine, which is mostly prescribed in outpatient settings and filling a prescription does not imply that it is taken as prescribed. Similarly, the variations in billing patterns for methadone necessitated the use of grace periods which might have introduced some misclassification. However, the consistency of findings using different definitions of treatment discontinuation provides reassurance that the impact of misclassification is likely to be small. Second, although we accounted for a broad range of potential confounders, residual confounding cannot be ruled out. Variables such as patient-level socioeconomic status, behavioral factors, and patient access to different types of care were not available in our data. Lastly, despite including more than 15 years of nationwide data, several factors contributed to a reduction in study size. Some state Medicaid programs, especially in the early years, did not cover MOUD (e.g., 20 states did not cover methadone in 2011–2013).³⁹ Although buprenorphine is primarily prescribed through routine outpatient practice, which is fully captured in prescription fill records, it can also be prescribed through opioid treatment programs (OTPs), which is only partially captured in our data. This group is expected to represent a relatively small proportion compared with those receiving buprenorphine through office-based practices, as OTP-based prescribing is less common for buprenorphine.^40,41 To ensure data accuracy and completeness for assessing treatment retention and ascertainment of confounding factors, we required continuous enrollment from 3 months before pregnancy to 1 month after delivery and excluded data from several states in select years due to data quality issues.^15,32 The generalizability of our findings should be viewed in light of these restrictions. In addition, the small number of events for some of the OUD- and mental health-related adverse outcomes during pregnancy and the postpartum period (e.g., opioid overdose, any drug overdose, and self-harm), precluded us from drawing definitive conclusions about the comparative effectiveness of methadone versus buprenorphine for these events.

Our study provides clinical and policy insights for OUD treatment during pregnancy and the postpartum period. As recommended by guidelines,¹² the choice between different MOUDs should be a shared decision between patients and physicians, taking into account multiple factors, including efficacy, safety, patient preference, and patient access to different types of care. Treatment retention is another factor to be considered. Importantly, the high risks of discontinuation for both buprenorphine and methadone underscore the critical need to identify and address modifiable clinical and health-services factors that can improve treatment retention in pregnant patients with OUD, and thus maternal and neonatal outcomes. Several key factors highlighted in prior literature include reduced access to MOUD treatment among pregnant patients facing disproportionate challenges (e.g., non-white patients), and the stigma and legal-related challenges faced during pregnancy and postpartum, such as reduced motivation to seek healthcare or treatment due to fear of losing child custody.^42–44 Moreover, it is possible that switching between buprenorphine-alone and buprenorphine/naloxone during pregnancy and the postpartum period may increase the risk of discontinuation. Cost, availability, and prior authorization requirements have been suggested as factors that complicate the process of switching between MOUDs.⁴⁵ These barriers should be minimized to avoid treatment gaps during transitions. The growing evidence supporting the safety of naloxone use during pregnancy^46–48 supports the guidance reflected in some of the newer guidelines that patients stable on buprenorphine/naloxone do not necessarily need to be switched to buprenorphine-alone after learning of the pregnancy^12,45 and suggests that broader adoption of this position might be warranted. As more recent, nationwide data become available, it will be important to evaluate whether retention in MOUD treatment is changing over time. It would provide insights into the effects of the shifts toward synthetic opioids, primarily fentanyl, which now are the leading cause of opioid overdose-related deaths;^49,50 as well as how recent updates to the evidence and recommendations for buprenorphine/naloxone use during pregnancy,^12,45–48 and other policy changes (e.g., the increased use of telemedicine and expanded allowances for take-home methadone supplies following the COVID-19 pandemic)⁵¹, affect MOUD treatment retention among the pregnant population.

Conclusions

The risks of MOUD discontinuation were high during pregnancy and the postpartum period. About a third of patients discontinued MOUD treatment during pregnancy, and by the end of the first year postpartum about half have discontinued. Pregnant patients initiating transmucosal buprenorphine during early pregnancy were more likely to discontinue treatment than those initiating methadone. Our findings highlight the importance of identifying and addressing barriers to treatment retention among pregnant patients with OUD.