FormalPara Key Summary Points

Why carry out this study?

 Erenumab, an anti-calcitonin gene-related peptide (CGRP) receptor monoclonal antibody (mAb), was approved by the US Food and Drug Administration in May 2018. Constipation with serious complications was added to the Warning and Precautions section in the erenumab Prescribing Information in October 2019 after events were observed during post-marketing surveillance.

 We aimed to assess and compare the risk of inpatient constipation, and, separately, inpatient constipation with serious complications, among patients with migraine treated with CGRP mAbs and standard of care antiepileptic drugs (AEDs).

What was learned from this study?

 Patients initiating erenumab had similar risk of inpatient constipation within 90 days of treatment initiation versus patients initiating other CGRP mAbs, and lower risk versus patients initiating AEDs.

 Findings from this study provide context to events observed in the post-marketing setting.

Introduction

Migraine is a neurological disorder characterized by recurrent headache attacks of moderate to severe pain. In 2018, the prevalence of migraine or severe headache was 21.0% in women and 10.7% in men in the USA [1]. Treatment options include medications for acute episodes (e.g., triptans) and preventive medications to reduce the frequency of migraines. Medications including antiepileptics, beta blockers, calcium channel blockers, antidepressants, and botulinum toxins are prescribed for prevention, but have primary indications for other conditions such as epilepsy, depression, and hypertension [2].

Therapies targeting the calcitonin gene-related peptide (CGRP) pathway have been developed for both acute and preventive treatment of migraine. One class, known as gepants, consists of oral, non-peptide, small molecule, CGRP receptor antagonists. The other class of medications is the anti-CGRP pathway monoclonal antibodies (referred to hereafter as CGRP mAbs) which were developed specifically for the prevention of migraine and are administered by injection or infusion. Four CGRP mAbs have been approved by the US Food and Drug Administration (FDA): erenumab-aooe (Aimovig®, Amgen Inc., Thousand Oaks, CA) on 17 May 2018, fremanezumab-vfrm (Ajovy®, Teva Pharmaceuticals USA, Parsippany, NJ) on 14 September 2018, galcanezumab-gnlm (Emgality®, Eli Lilly and Company, Indianapolis, IN) on 27 September 2018, and eptinezumab-jjmr (Vyepti®, Lundbeck Seattle BioPharmaceuticals Inc., Bothell, WA) on 21 February 2020. Erenumab is a mAb that targets the CGRP receptor, whereas fremanezumab, galcanezumab, and eptinezumab are mAbs that target the CGRP ligand.

Among patients included in three placebo-controlled clinical studies of erenumab [3,4,5], the incidence of constipation was 1% with placebo, 1% with 70 mg erenumab, and 3% with 140 mg erenumab during the 12-week double-blinded treatment phase, as described in the Adverse Reactions section of the Prescribing Information upon FDA approval [6]. In the post-marketing setting, constipation with serious complications was reported following the use of erenumab and was added to the Prescribing Information for erenumab in October 2019 [6].

The purpose of this retrospective observational study was to estimate and compare the risk of inpatient constipation (constipation recorded during an inpatient stay or emergency department (ED) visit) among patients with migraine following initiation of treatment with a preventive migraine medication. Multiple risk windows for assessment of inpatient constipation were evaluated, with the 90 days following treatment initiation specified a priori as the primary window of interest. Additionally, the risk of serious complications within 30 days following inpatient constipation was assessed. The medications examined included erenumab, other CGRP mAbs, and standard of care antiepileptic drugs (AEDs). We hypothesized that the risk of inpatient constipation among initiators of erenumab would be similar to the risk among initiators of other CGRP mAbs and initiators of AEDs.

Methods

Data Source

In this retrospective cohort study, the study population was drawn from Optum’s Electronic Health Record (EHR) Research Database, a de-identified patient-level database that integrates multiple electronic medical record (EMR) data systems with medical claims, prescription, and practice management data. The database incorporates relevant clinical data on patients from both ambulatory and inpatient settings, including medical records, laboratory results, and drug prescription and administration data, as recorded during routine clinical practice.

For 2019, data relating to 32 million patients with at least one medical encounter were available. The data are collected from more than 92,000 providers and 195 hospitals representing 50 EMR-based provider/hospital networks across the USA. The population captured in Optum’s EHR Research Database is geographically diverse and not specific to health insurance so that patients with commercial health insurance are present along with patients with coverage through Medicare, Medicaid, and even patients with no health insurance.

The database is certified as de-identified by an independent statistical expert following Health Insurance Portability and Accountability Act statistical de-identification rules; the study protocol was exempt from institutional review board review.

Study Population

Patients with a prescription order for erenumab, other CGRP mAbs (fremanezumab, galcanezumab, eptinezumab), and AEDs (carbamazepine, gabapentin, topiramate, valproate sodium/valproic acid/divalproex sodium, zonisamide) between 17 May 2018 and 31 March 2020 were identified using National Drug Codes (NDCs). To restrict to new users, only the first prescription order per patient and per medication was identified during the study period and assessed for the following criteria:

  • At least 18 years of age on the prescription order date

  • At least two diagnosis codes for migraine (International Classification of Diseases, 10th revision (ICD-10) code G43.*) on two different days, or at least one diagnosis code for migraine and one prescription order for an acute migraine treatment (triptan or ergot), in the 12 months prior to and including the prescription order date (to ensure cohort members were patients with migraine, especially those initiating AEDs, which have other indications)

  • At least two outpatient visits prior to the prescription order date, including one visit at least 12 months prior to the prescription order date (to establish a 12-month baseline period for assessment of clinical covariates)

  • No prescription order for any CGRP mAb during the prior 12 months

  • For the AED cohort only, no prescription order for any of the five antiepileptic medications during the 12 months prior to the prescription order date

The index date was set as the date of the earliest prescription order that met all the aforementioned criteria. Because it was expected that patients would have started a CGRP mAb after attempting treatment with an AED, the AED cohort was selected from the remaining population of migraine preventive treatment users after the erenumab and other CGRP mAb cohorts were formed.

Ascertainment of Covariates

The baseline period for covariate assessment was the 12 months prior to and including the index date. Demographic variables were extracted directly from the EHR database. Clinical variables were assessed with ICD-10 diagnosis codes, NDCs, and procedure codes derived from the structured data tables in the database. They included comorbidities related to migraine, preventive migraine treatments, drugs that may cause or treat constipation, gastrointestinal disorders, conditions related to constipation, and constipation history. Personal and family history of constipation and enemas used for constipation treatment (gastrografin, glycerin/Fleet) were identified using semi-structured data (i.e., data derived from natural language processing of clinical notes) within the EHR database.

Propensity Score Modeling and Matching

Two separate logistic regression models were used to estimate propensity scores (PSs) predicting initiation of erenumab compared to initiation of (1) other CGRP mAbs and (2) AEDs. PSs were estimated on the basis of a priori potential confounders and risk factors for constipation. The variables included in the PS model are listed in Tables 1 and 2.

Table 1 Risk factors for constipation in erenumab and other CGRP monoclonal antibody initiators in the 12-month baseline period prior to the index date, pre- and post-propensity score matching
Full size table
Table 2 Risk factors for constipation in erenumab and standard of care antiepileptic drug initiators in the 12-month baseline period prior to the index date, pre- and post-propensity score matching
Full size table

Each erenumab initiator was matched on the basis of PS to one initiator in each comparator cohort (1:1). Initiators were matched without replacement, using a greedy matching algorithm with a variable caliper that allowed for a maximum PS difference of 0.1 [7, 8].

Identification of Outcomes

Inpatient constipation was identified on the basis of ICD-10 diagnosis codes for constipation (K59.00–K59.09) that were recorded during an ED or inpatient stay. Inpatient constipation was assessed starting on the day after the index date through the earliest occurrence of switching of migraine preventive therapy or end of the study period (31 March 2020). Only the first (incident) constipation event occurring during the follow-up period was included in the analysis.

Serious complications of inpatient constipation were assessed within 30 days following the inpatient constipation event. Serious complications were identified by the presence of at least one ICD-10 diagnosis code for fecal impaction, intestinal obstructions, anal fissures, fistulas, or related conditions identified in an ED or inpatient setting (Table S1 in electronic supplementary material).

Statistical Analysis

All analyses were conducted using SAS version 9.4 (SAS Institute Inc, Cary, NC). Patient characteristics were summarized for each of the cohorts before and after PS matching using frequencies and percentages for categorical variables and means and standard deviations for continuous variables. The PS matching process resulted in some unmatched erenumab cohort members. For transparency, characteristics of the unmatched erenumab cohort members were described. Standardized differences were examined to assess the balance of each risk factor between the matched erenumab and comparator cohorts. Covariates with an absolute standardized difference no greater than 0.1 were considered balanced [9, 10].

Risk (i.e., incidence proportion) and corresponding 95% confidence intervals (CIs) of inpatient constipation were estimated in the three cohorts by dividing the total number of patients with an event observed during follow-up by the number of cohort members at risk at the start of follow-up. Risk was calculated among the overall cohorts and the PS-matched cohorts. The risk of serious complications of inpatient constipation was estimated in the same manner.

The risk windows for outcome assessment included 30, 60, and 90 days following the index date, and all available follow-up, with the 90-day window specified a priori as the primary risk window. The accrual period for initiators was adjusted to ensure the patients included in the analysis for each risk window had the requisite amount of follow-up time. For each risk window, patients who initiated treatment during the following time periods were included:

  • 30-day risk window: 17 May 2018 through 29 February 2020

  • 60-day risk window: 17 May 2018 through 31 January 2020

  • 90-day risk window: 17 May 2018 through 31 December 2019

  • All available follow-up: 17 May 2018 through 31 March 2020

Odds ratios (ORs) and corresponding 95% CIs were estimated using logistic regression models to compare the risk of inpatient constipation in the matched erenumab cohort to the other CGRP mAb and, separately, to the AED cohorts. The OR approximates the estimate of interest, the risk ratio, because inpatient constipation is a rare outcome [11]. The outcome models were not adjusted for covariates as none were imbalanced following PS matching.

As a sensitivity analysis, the risk of inpatient constipation was calculated within the study cohorts after stratifying on the presence of epilepsy diagnosis and the use of constipation-causing medications during the baseline period. Additionally, among erenumab initiators only, the risk of inpatient constipation by prior use of AEDs was assessed. Finally, to ensure that amount of available follow-up time among initiators of erenumab and initiators of other CGRP mAbs was comparable, the risk of inpatient constipation was calculated among the subset of PS-matched initiators whose index date was on or after 1 January 2019.

Results

Cohort Formation and Baseline Characteristics

Between 17 May 2018 and 31 March 2020, we identified 29,372 patients whose earliest prescription order was for erenumab, 21,962 patients whose earliest prescription was for other CGRP mAbs, and 243,183 patients whose earliest prescription was for AEDs (Fig. 1). Following application of all study criteria, 17,902 patients were included in the erenumab initiator cohort, 13,404 patients were included in the other CGRP mAb initiator cohort, and 49,497 patients were included in the AED initiator cohort. Among patients in the other CGRP mAb cohort, 63.4% had a prescription order for galcanezumab on the index date, 36.3% had a prescription for fremanezumab, 0.3% had prescriptions for both galcanezumab and fremanezumab, and 0% had a prescription for eptinezumab, which was approved on 21 February 2020 (data not shown).

Fig. 1
figure 1

Formation of erenumab, other CGRP monoclonal antibody, and standard of care antiepileptic drug initiator cohortsa. AED antiepileptic drug, CGRP calcitonin gene-related peptide, mAbs monoclonal antibodies. aTo identify initiators, only the earliest prescription order during the study period was assessed for cohort eligibility. bPatients were also required to have known geographic region

Full size image

Tables 1 and 2 present the baseline characteristics of the erenumab, other CGRP mAb, and AED initiator cohorts pre- and post-matching. Pre-matching, mean age was similar across cohorts, ranging from 45.2 years in the other CGRP mAb and AED cohorts to 45.9 years in the erenumab cohort. The majority of patients in each cohort were female: 87.5% of erenumab initiators, 88.2% of other CGRP mAb initiators, and 86.4% of AED initiators. Pre-matching, inpatient constipation during the 12-month baseline period was observed among 1.9% of patients in the erenumab cohort, 1.7% of patients in the other CGRP mAb cohort, and 2.8% of patients in the AED cohort.

For the erenumab–other CGRP mAb comparison, 13,200 initiators in each cohort were PS-matched (Table 1); 4702 erenumab initiators were unmatched. For the erenumab-AED comparison, 15,441 initiators in each cohort were matched (Table 2); 2461 erenumab initiators were unmatched. For both comparisons, there were no risk factors with an absolute standardized difference greater than 0.1 in the PS-matched cohorts, indicating the matched cohorts were well balanced. The distribution of PSs before and after PS matching are shown in Figs. S1a and S1b for the erenumab and other CGRP mAb initiator cohorts and Figs. S2a and S2b for the erenumab and AED initiator cohorts (electronic supplementary material).

Risk of Inpatient Constipation and Serious Complications of Inpatient Constipation

Within the 90-day risk window following the index date, we observed 84 inpatient constipation events among 15,983 erenumab initiators, 50 events among 11,345 other CGRP mAb initiators, and 398 events among 43,810 AED initiators before PS matching (Table 3). The corresponding risks were 0.53% (95% CI 0.42–0.65) for erenumab, 0.44% (95% CI 0.33–0.58) for other CGRP mAb, and 0.91% (95% CI 0.82–1.00) for AED initiators.

Table 3 Risk of inpatient constipation within a 90-day risk window among erenumab, other CGRP monoclonal antibody, and standard of care antiepileptic drug initiators, pre- and post-propensity score matching
Full size table

For the erenumab–other CGRP mAb comparison, results were similar after PS matching. The risk of inpatient constipation was 0.46% (95% CI 0.35–0.60) among matched erenumab initiators and 0.44% (95% CI 0.33–0.58) among matched other CGRP mAb initiators, with a corresponding OR of 1.06 (95% CI 0.72–1.55) (Table 3). For the erenumab–AED comparison, the risk of inpatient constipation was 0.53% (95% CI 0.42–0.66) among matched erenumab initiators and 0.76% (95% CI 0.62–0.92) among matched AED initiators, with a corresponding OR of 0.69 (95% CI 0.51–0.94).

Within 30 days after the inpatient constipation event, the risks of serious complications of inpatient constipation were 0.05% (95% CI 0.03–0.10), 0.04% (95% CI 0.02–0.10), and 0.10% (95% CI 0.07–0.13) among the erenumab, other CGRP mAb, and AED initiators, respectively (Table 4). After PS matching, the risk of serious complications remained low in each cohort.

Table 4 Risk of serious complications of inpatient constipation within a 90-day risk window among erenumab, other CGRP monoclonal antibody, and standard of care antiepileptic drug initiators, pre- and post-propensity score matching
Full size table

Table S2 (electronic supplementary material) presents the risk of inpatient constipation among the erenumab, other CGRP mAb, and AED initiators within different risk windows following the index date. Within a 30-day risk and 60-day risk window, the risk was similar among matched erenumab and other CGRP mAb initiators. When assessed during the all available follow-up risk window, the risk was 1.52% (95% CI 1.32–1.74) among matched erenumab initiators and 1.08% (95% CI 0.92–1.27) among matched other CGRP mAb initiators, with an OR 1.40 (95% CI 1.13–1.74). For the erenumab–AED comparison, the risk among matched erenumab initiators was lower than that among matched AED initiators within the 30-day, 60-day, and all available follow-up risk windows. Like the 90-day risk window, there were few serious complications of constipation observed within the erenumab, other CGRP mAb, and AED initiator cohorts for the other risk windows (Table S3, electronic supplementary material).

Table S4 (electronic supplementary material) provides the risk of inpatient constipation within 90 days following the index date among the erenumab, other CGRP mAb, and AED initiators stratified by presence of epilepsy. For all cohorts, the risk of inpatient constipation was higher among initiators with an epilepsy diagnosis during baseline compared to initiators without a diagnosis. However, the OR for the erenumab–other CGRP mAbs comparison was similar among those with an epilepsy diagnosis and those without a diagnosis. Similar results were observed for the erenumab–AEDs comparison; the ORs among patients with and without epilepsy were 0.76 (95% CI 0.35–1.62) and 0.67 (95% CI 0.48, 0.92), respectively.

The risk of inpatient constipation within 90 days following the index date stratified by baseline medication use is presented in Table S5 (erenumab and other CGRP mAb initiators) and Table S6 (erenumab and AED initiators) in the electronic supplementary material. The risk of inpatient constipation was higher among erenumab, other CGRP mAb, and AED initiators with baseline use of opioids, anticholinergic medications, non-steroidal anti-inflammatory drugs (NSAIDs), cation-containing agents, and serotonin (5-HT3) receptor antagonists compared to initiators without baseline use of these medications. Additionally, the risk of inpatient constipation was higher among erenumab initiators with AED use during baseline, compared to erenumab initiators without AED use during baseline.

Table S7 (electronic supplementary material) presents the risk of inpatient constipation and corresponding ORs within each risk window among the subset of erenumab and other CGRP mAb initiators whose index date was on or after 1 January 2019. Within a 90-day risk window, the OR for matched erenumab versus other CGRP mAb initiators was 0.90 (95% CI 0.57–1.40). For all available follow-up time, the OR was 1.14 (95% CI 0.87–1.49).

Discussion

In the post-marketing setting, inpatient constipation with serious complications was reported following the use of erenumab. As such, this study assessed the risk of constipation and serious complications among patients with migraine who initiated preventive treatment with erenumab, other CGRP mAbs, or standard of care AEDs in the inpatient setting only; constipation associated with outpatient visits was not assessed. Since antiepileptics are commonly used as migraine preventive agents, they were chosen as the standard of care therapy for the second comparator cohort to avoid drug classes known to be associated with an increased risk of constipation (e.g., antihypertensives, antidepressants).

The risk of inpatient constipation in the 90 days following treatment initiation was similar for the erenumab and other CGRP mAbs cohorts, while a lower risk was observed among erenumab initiators compared to AED initiators. Moreover, inpatient constipation risk was higher among AED initiators than among initiators of erenumab or other CGRP mAbs, even after PS matching. A possible explanation for this finding may be that carbamazepine and valproic acid were among the antiepileptic medications included in the comparator cohort; constipation is a side effect of both treatments [12].

There were few serious complications of constipation observed in this study; the risk of serious complications of constipation was low overall and similar in the erenumab and other CGRP mAb cohorts, but slightly higher in the AED cohort.

The risk of inpatient constipation among all initiators of erenumab in this study was 0.53% within 90 days following treatment initiation. This estimate is similar to the incidence reported in a retrospective cohort study conducted within the MarketScan® Research Databases [13]. Among patients with migraine initiating various acute and preventive migraine treatments, the incidence of serious constipation (i.e., constipation claim in an ED or inpatient setting) was 0.63% [13]. However, the risk observed in this study was lower than that observed among patients in the erenumab clinical studies [3,4,5], where incidence of any constipation during the first 3 months was 1% with placebo, 1% with 70 mg erenumab, and 3% with 140 mg erenumab [6]. Other studies conducted using real-world data have also reported higher incidence of any constipation (13.5–23.9%) among patients treated with erenumab, although most cases were mild, suggesting that constipation may be a frequent but minor effect of erenumab treatment [14,15,16]. As our study assessed inpatient constipation only, it is expected that the risk would be lower than studies that identified any constipation, but may impact the generalizability of the results.

We observed a higher risk of inpatient constipation among erenumab, other CGRP mAb, and AED initiators with constipation risk factors during baseline, including an epilepsy diagnosis and use of medications known to be associated with constipation, such as opioids, anticholinergics, and 5-HT3 receptor antagonists. This suggests that some cases of inpatient constipation may be partially attributable to these factors rather than to use of the CGRP mAbs alone.

In this study, we did not assess gepants which are small molecules that also target the CGRP pathway to treat migraine. Gepants were not approved by the FDA until late in the study period (December 2019 for ubrogepant and February 2020 for rimegepant) or after the study was completed (September 2021 for atogepant). Nonetheless, it is possible that some patients included in this study received a gepant prior to the index date or before the end of the study period.

An advantage of conducting this analysis within an EHR database versus a claims database is that patient assistance programs sponsored by pharmaceutical manufacturers are unlikely to have impacted the identification of treatment initiators in this study. While EHR data are valuable for the examination of clinical outcomes and treatment patterns, EHR databases have certain inherent limitations because the data are collected for the purpose of clinical patient management, not research. The presence of a diagnosis code may not represent the true occurrence of disease, as the diagnosis may be incorrectly coded or included as rule-out criteria rather than actual disease. Furthermore, a diagnosis code for inpatient constipation in this study may have included events for which constipation was the reason for the admission, present on admission, or developed during the hospital stay. Additionally, the prescription data represent the intent of the prescriber through the written prescription for a medication, and do not indicate that a medication was filled, consumed, or taken as prescribed.

It is possible that patients in the AED cohort took their index medication for an indication other than migraine. To ensure these cohort members were patients with migraine, a combination of two migraine diagnosis codes and/or prescriptions for migraine treatments were required in the 12-month baseline period. Analyses stratified by the presence of an epilepsy diagnosis code during the baseline period were also conducted; the OR of inpatient constipation for erenumab relative to AEDs was similar among initiators with a baseline epilepsy diagnosis and initiators without a baseline epilepsy diagnosis.

As is true for most clinical record-keeping systems, it is not possible to directly determine the completeness of data capture during baseline and follow-up periods within Optum’s EHR database as some patients may receive only a portion of their care from a provider included in the database. Furthermore, we cannot confirm when a patient is lost to follow-up. This contrasts with claims databases, where baseline and follow-up are defined on the basis of dates of health plan enrollment, ensuring that capture of clinical encounters in the database during those periods is relatively complete. In this study, visit dates were available in the EHR database to determine when events of interest occurred. A proxy could have been used to define the end of follow-up, such as the date of last encounter, but this approach would enable sicker patients with more frequent medical visits to contribute more follow-up time than healthier patients with fewer medical visits. To avoid this form of selection bias, the risk (incidence proportion), which was calculated using the number of cohort members as the denominator, was estimated rather than the incidence rate, which is calculated using the person-time at risk as the denominator. Given that the primary objective was to identify inpatient constipation events within a relatively short period following drug exposure (i.e., 90 days), the incidence proportion could serve as a proxy of the cumulative incidence.

In this study, multiple risk windows for outcome assessment were evaluated, with the 90-day risk window specified a priori as the primary risk window of interest. When initiators of erenumab were compared to initiators of other CGRP mAbs, the risk of inpatient constipation was similar in the two cohorts for the 30-day, 60-day, and 90-day risk windows. However, for all available follow-up, risk of inpatient constipation was higher among erenumab initiators compared to the other CGRP mAb initiators. In contrast, when the erenumab cohort was compared to the AED cohort, the risk of inpatient constipation was consistently lower in the erenumab cohort across all risk windows. The consistency in the ORs obtained by assessing occurrence of outcomes in risk windows of varying lengths provides assurance on the robustness of the estimates observed in this study.

The start of the study period was 17 May 2018, the date that erenumab was approved in the USA. However, fremanezumab and galcanezumab were not approved until September 2018. Consequently, the follow-up period was longer for erenumab compared to other CGRP mAb initiators, providing greater opportunity for the occurrence of outcomes among erenumab initiators. Indeed, when the main results were compared to those from the sensitivity analysis among the subset of matched initiators whose index date was on or after 1 January 2019, the ORs were attenuated in the sensitivity analysis. This finding suggests that a longer period of available follow-up among the erenumab compared to other CGRP mAb initiators may have contributed to the higher risk of inpatient constipation observed in this cohort, particularly during all available follow-up in the main analysis.

Although adjustment for confounding was implemented through PS modeling and matching, residual confounding is possible. As health plan coverage for erenumab may require prior authorization or a “step-through” therapy (i.e., failure of standard of care medications), the erenumab initiators may have been different from the AED initiators with respect to migraine severity. This difference could have resulted in biased measures of association if migraine severity is associated with constipation [17, 18]. Following PS matching, the study cohorts were found to be comparable with respect to the measured confounders. Nonetheless, migraine severity was not measured directly, although proxies for severity (i.e., use of migraine preventive agents, comorbidities related to migraine) were included in the PS model. Compared to initiators of AEDs, initiators of other CGRP mAbs were likely more comparable to initiators of erenumab with respect to migraine severity, as indicated by the prevalence of proxies for severity pre-matching. The findings observed in this study population, consisting predominantly of women who likely experienced moderate–severe migraine, may not be generalizable to the overall population of patients with migraine, despite the significance of migraine severity as a confounder.

Conclusion

The risk of inpatient constipation within 90 days following treatment initiation among patients who initiated erenumab was similar to the risk among patients who initiated other CGRP mAbs and lower than the risk among patients who initiated AEDs. Additionally, the risk of serious complications of inpatient constipation overall was low and comparable in the erenumab and other CGRP mAb cohorts, but slightly higher in the AED cohort. However, as few serious complications of inpatient constipation were observed and the 95% CI for the risks were wide, results for this outcome should be interpreted with caution. Nonetheless, these findings provide context to events observed in real-world post-marketing surveillance data.