Chronic Opioid Use Is Associated With Early Biologic Discontinuation in Inflammatory Bowel Disease

Christian Rhudy; Courtney L. Perry; Michael Singleton; Jeffery Talbert; Terrence A. Barrett

Disclosures

Aliment Pharmacol Ther. 2021;53(6):704-711. 

In This Article

Results

Patient Characteristics

The final study cohort comprised of 16 624 individuals with IBD who were initiated on biologic therapy during the study period. In this cohort, we identified 9013 (54.2%) individuals who received an opioid prescription, with 1768 (10.6%) meeting criteria for chronic opioid use. Substance use disorder was diagnosed in 26.9% of the chronic opioid use cohort, and 15.2% received opioid agonist therapy. Subjects with chronic opioid use were observed to have greater periods of continuous insurance coverage than others in the study population, which allowed for longer follow-up during this study period. Individuals with chronic opioid use had a median follow-up time of 1443 days compared to 1270 days in those without a diagnosis of chronic opioid use (P < 0.0001; Table 1). A significantly greater proportion of IBD patients with chronic opioid use had Crohn's disease as compared to those without (87% vs 84.2%, P = 0.002), whereas no significant difference between the cohorts was observed in the incidence of ulcerative colitis and indeterminate colitis. For this reason, disease subtype was included in the multivariable model. However, other results were not classified by disease subtype in order to simplify reporting.

Significant differences in patient demographics were noted among those with chronic opioid use, including sex, age and geographic region. Patients with chronic opioid use were predominantly female (57.4% vs 50.8%) and older (mean age of 45 vs 37). The majority of patients with chronic opioid use were in the age range of 45–65 years old (55.1%), while just 35.7% of IBD patients without chronic opioid use fell into this same range (P < 0.0001). Patients with chronic opioid use were more likely to reside in the 'Southern' region than those without chronic opioid use (P < 0.0001), whereas patients without chronic opioid use were more likely to reside in the 'Northeast.'

Chronic Opioid use and Biologic Therapy Usage

IBD patients in the chronic opioid use cohort were significantly more likely to utilise more than one biologic agent during the study period (P < 0.0001; Table 2). Subjects with chronic opioid use utilised a mean 1.5 different biologics, as compared to 1.37 biologics in the reference IBD cohort (P < 0.0001). Individuals with chronic opioid use were significantly more likely to utilise adalimumab (P < 0.0001), golimumab (P = 0.0003), certolizumab (P < 0.0001) and ustekinumab (P = 0.029) than those in the reference IBD cohort. Furthermore, individuals with chronic opioid use were less likely to receive infliximab (P = 0.0002). There was no significant difference in vedolizumab use between groups.

In addition to a higher number of unique biologic agents used over the study period, we found that IBD patients with chronic opioid use had decreased duration of persistence to biologic therapy than their counterparts without chronic opioid use (Table 3). A significantly greater proportion of patients with chronic opioid use had decreased persistence to biologic therapy when compared to the IBD reference group (P < 0.0001), as evidenced by the low percentage of biologic therapy persistence observed in the chronic opioid use group (16.2%) compared to the IBD reference group (33.5%) at the end of the study period (P <0.0001) (Figure 1). Treatment duration on the first and second biologic agents prescribed for IBD was significantly shorter in patients with chronic opioid use compared to those without (P < 0.0001). However, no significant difference was observed between groups by the third, fourth or fifth lines of therapy.

Figure 1.

Risk of discontinuation of biologic therapy with 95% confidence interval

Comorbidities and Disease Severity

Healthcare utilization, including emergency department visits, hospital admissions, abdominal surgeries, mean number of endoscopies and steroid usage, was observed to be significantly higher in the chronic opioid use cohort (P < 0.0001) (Table S1). The rates of common comorbidities and conditions, including fistula, malnutrition, mental disorder, cancer, chronic pain and arthritis diagnoses, were all significantly higher in the chronic opioid use cohort compared to the reference IBD patients. The average Charlson comorbidity index was significantly higher in chronic opioid use patients compared to the reference IBD patients (1.17 vs 0.69; P < 0.0001). The chronic opioid use cohort experienced more comorbid conditions and disease complications, while also utilising healthcare resources to a greater extent than the reference cohort.

Predictive Modelling

Univariable survival curves by chronic opioid use status were significantly different (P < 0.0001). Subjects with chronic opioid use were 56% more likely to be non-persistent (hazard ratio 1.56, 95% CI [1.48–1.65]). Among other categorical variables, only Crohn's disease failed to show a significant univariable association with risk of non-persistence (hazard ratio 0.98, 95% CI [0.93–1.03]) (Table S2). Among continuous covariates, risk of discontinuation appeared to increase linearly in both the number of emergency department visits (hazard ratio 1.06, 95% CI [1.05–1.07]) and outpatient steroid use (hazard ratio 1.018, 95% CI [1.015–1.022]) as measured in 1000 mg equivalents prescribed. Non-linear relationships were observed with the number of inpatient admissions and inpatient steroid exposure, and were taken into account for the final multivariable model Table S3.

After variable selection and assessment of the proportional hazard assumption, the main effects model included chronic opioid use, dichotomous age (<50 vs ≥50), gender, arthritis, mental disorder, abdominal surgery, inflammatory bowel disease type, number of inpatient admissions and inpatient steroid exposure (Table 4). Subjects with chronic opioid use were 23% more likely to be non-persistent to biologic therapy, after adjusting for the other variables included in the final multivariable model (hazard ratio 1.23, 95% CI [1.16–1.31]). Diagnostics indicated that the final model was stable and well-calibrated (P-values in all 10 risk score groups were well above 0.05). Concordance was c = 0.60, indicating moderate predictive accuracy.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....