Vedolizumab is associated with a lower risk of serious infections than anti-TNF agents in older adults

Abstract

Background:

Despite the rising numbers of older adults with inflammatory bowel diseases (IBD), there are few studies regarding safety and effectiveness of IBD treatments in older adults.

Aims:

To compare the safety and effectiveness of anti-tumor necrosis factor (TNF)-α agents and vedolizumab in older adults with IBD.

Methods:

We conducted a retrospective cohort study using an active comparator, new user design for adults ≥65 years with IBD initiating anti-TNF-α agents and vedolizumab in the Medicare-claims database from 2014–2017. The primary safety outcome was infection-related hospitalization (excluding intra-abdominal and perianal abscesses). Co-primary outcomes to estimate effectiveness were IBD-related hospitalization, IBD-related surgery and new corticosteroid use ≥60 days after biologic initiation. We performed propensity score weighting to control for confounding and estimated adjusted hazard ratios (aHR) and 95% confidence intervals (CIs) using standardized morbidity ratio (SMR)-weighted variables.

Results:

We identified 1,152 anti-TNF-α new users and 480 vedolizumab new users. Median age was 71 years in both cohorts and 11% were ≥80 years. Crohn’s disease patients comprised 54% of the anti-TNF-α cohort and 57% of the vedolizumab cohort. There was no significant difference in demographics, healthcare utilization and frailty in both cohorts. Greater than half of both cohorts had a Charlson co-morbidity index of ≥2. Vedolizumab users had a decreased risk of infection-related hospitalization (aHR: 0.47, 95% CI: 0.25 – 0.86). There was no significant difference in the outcomes approximating effectiveness.

Conclusion:

Older IBD patients treated with vedolizumab had a lower risk of infection-related hospitalization compared with those initiating anti-TNFs. We observed no difference in effectiveness defined by hospitalizations, surgery and new corticosteroid use.

INTRODUCTION

Inflammatory bowel diseases (IBD), comprising Crohn’s disease (CD) and ulcerative colitis (UC), have a bimodal peak of incidence with the second peak occurring in the seventh decade of life.¹ At least 20% of incident IBD diagnoses occur in adults ≥60 years.² The aging population, decreasing fatality of IBD, improved treatments and rise in IBD incidence and prevalence result in a prominent increase in older adults with IBD.^{3, 4} In 2015 26% of Americans with IBD were ≥65 years.⁵ A modeling study projects that older adults with IBD will increase >200% by 2030.⁶ Despite growing numbers, there are few dedicated studies regarding safety and effectiveness of medications used to treat IBD in older adults.

Effective immunosuppression with biologic agents is the main stay of treatment for moderate-to-severe IBD.^{7, 8} Treatment options for IBD have proliferated resulting in the need to position therapies.^{9, 10} The studies providing evidence to position biologic agents, do not focus on older adults.^11–13 The only comparative effectiveness study of biologic agents focused on older adults is very limited in generalizability.¹⁴

Older adults treated for IBD are at increased risk for complications, particularly infections.^{15, 16} Older adults also may have different responses to medications.¹⁷ Therefore, there is an unmet need to examine the comparative safety of biologic agents in a large and unselected cohort of older adults. We aimed to determine safety and effectiveness of biologic therapies in older adults with IBD.

METHODS

We conducted a retrospective cohort study using an active comparator, new-user design for adults ≥65 years with IBD initiating anti-tumor necrosis factor (TNF)-α agents or vedolizumab. The active comparator, new-user design is designed to minimize biases.¹⁸

Data Source

We used a 20% random sample from the 50-state Medicare claims database. Please see supplemental methods for details.

Study Population

We included individuals ≥65 years. We defined CD and UC per prior studies of IBD leveraging claims data.¹⁹ (Details are in supplemental methods)

Study Design

We stipulated a minimum of 15-months of continuous enrollment prior to drug initiation to ensure no prior and to ensure adequate time to assess baseline covariates. The study was from May 2014 to December 2017, the point of latest data availability. The primary exposure was ≥1 drug code for an anti-TNF-α agent or vedolizumab.

Patients who had ≥2 International Classification of Disease (ICD-9/ 10) codes for rheumatoid arthritis, ankylosing spondylitis, plaque psoriasis or psoriatic arthritis were excluded to ensure that the anti-TNF-α was prescribed for IBD. For the comparative safety analysis, we additionally excluded anyone who had ≥2 ICD codes for a non-dermatologic malignancy as this may bias infectious risk. Additional details are in the supplemental methods.

Covariates

Pertinent covariates were assessed using all available data prior to biologic initiation. Covariates included age, sex, race, other IBD- medications before and concomitant with (+/− 30 days) biologic initiation, presence of a pre-biologic infections, health care utilization, hospitalization and IBD surgeries.

We accounted for co-morbidity using each of the 17 individual components of the Charlson co-morbidity index (CCI).²⁰ Frailty is increasingly recognized as a construct affecting outcomes in IBD.²¹ Therefore, we included a variable estimating frailty at baseline.²²

Outcomes

The co-primary outcomes estimated safety and effectiveness. The safety outcome was a hospitalization for an infection (defined by the primary ICD code for the hospitalization per a landmark publication).²³ This did not include intra-abdominal or perianal infections. Adverse outcomes for the effectiveness analysis were IBD-related hospitalizations (defined by the primary diagnosis code for the hospitalization), IBD-related surgery (CPT codes in Supplementary Appendix) or new corticosteroid initiation ≥60 days after biologic initiation for those who did not use steroids at baseline. These outcomes have been previously defined as surrogates of clinical effectiveness in IBD studies in administrative claims databases.^{24, 25} A secondary outcome was all-cause hospitalization.

Analysis

We compared hospitalization for serious infections, IBD-related hospitalization and IBD-related surgery between the two study arms: vedolizumab and anti-TNF-α agents. For the primary analysis, follow-up ended when a patient dis-enrolled from Medicare coverage or the end of the study period. We ran sensitivity analyses censoring patients at the time of treatment discontinuation, defined as the absence of drug prescription for four months or when the patient switched to comparator treatment and a second sensitivity analysis using all available look back period data. We used all covariates to estimate propensity scores for receiving vedolizumab compared to anti-TNF agents. In order to compare the estimates across the two treatment arms, we standardized the covariate distribution of anti-TNF users to that of the vedolizumab users by applying standardized morbidity ratio weights (SMRW), thereby, allowing us to estimate the average treatment effect in the vedolizumab cohort.²⁶ Covariate balance was assessed using standardized mean differences (SMD). We calculated crude incidence rates (cIR) and estimated adjusted hazard ratios (aHR) with 95% confidence intervals (CI) using SMR-weighted Cox proportional hazard models with a robust variance estimator to account for weighting. We also conducted a sensitivity analysis excluding those who were on concomitant immunomodulator therapy to ensure that the outcomes measured were reflective of the biologic agent alone. All analyses were performed using SAS V9.4.

RESULTS

We identified 1,152 new-users of anti-TNF-α agents and 480 new-users of vedolizumab. The median age was 71 years in both cohorts and 11% were ≥80 years. Sex and race distributions parallel known disease demographics. Patients with CD comprised 54% of anti-TNF-α new-users and 57% of vedolizumab new-users. See Table 1 for details (and Supplementary Table 1 for weighted cohorts).

Table 1:

Characteristics of inflammatory bowel disease (IBD) patients ≥65 years initiating anti-tumor necrosis factor (TNF)-α or vedolizumab in the United States Medicare Claims Database

Characteristic	Anti-TNF (n=1,152)	Vedolizumab (n=480)
Median Age (IQR) in years	71 (68 – 76)	71 (68 – 76)
Age Group
% 66–69 years	37	35
% 70–79 years	52	53
% ≥80 years	11	12
% Female	59	55
% White	93	91
% Crohn’s Disease	54	57
Baseline Charlson Co-Morbidity Index
% 0	24	26
% 1	21	24
% ≥2	55	50
Median Predicted Probability of Frailty (IQR)	4 (2 – 7)	4 (2 – 7)
# of Baseline Office Visits
% 0 – 10	29	25
% 11 – 20	43	42
% ≥21	29	33
# of Baseline Emergency Room Visits
% 0	44	50
% 1	19	21
% ≥2	37	29
% with Hospitalization at Baseline	43	34
% with IBD-related hospitalization	20	8
% infection-related hospitalization	6	5
% with IBD-related surgery at Baseline	4	4
Baseline Non-Biologic IBD Medications
% Budesonide	27	27
% Systemic Corticosteroid	61	53
% Mesalamine	61	42
% Immunomodulator	29	34
Concomitant Non-Biologic IBD Medications
% Budesonide	11	10
% Systemic Corticosteroid	38	30
% Immunomodulator	18	16

IQR: Inter Quartile Range

Baseline: the 6-month time frame prior to biologic initiation

Concomitant: at the time of biologic initiation

Over half the patients in both cohorts had a CCI of ≥2 and there was no difference in frailty as well. There were also no significant differences in healthcare utilization in the 12 months prior to biologic initiation in both cohorts. There were no notable differences between those with CD and UC in CCI, frailty or healthcare utilization.

In the 12 months prior to biologic initiation, patients who initiated anti-TNF-α agents were more likely to have been treated with a mesalamine agent (62% versus 42%) and corticosteroid (68% versus 57%) compared with patients who initiated vedolizumab. Baseline mesalamine and corticosteroid use was higher in UC patients than CD patients. See table 1 for details. All differences in baseline covariates were balanced after weighting, minimizing confounding by these covariates.

At the time of biologic initiation, patients who initiated anti-TNF-α agents were more likely to be concomitantly treated with corticosteroids (44% versus 33%) compared with patients who initiated vedolizumab. Concomitant corticosteroid use was more common in UC than CD. In this cohort of anti-TNF-α users, 18% had a concomitant prescription for a thiopurine and 4% had a concomitant prescription for methotrexate. In the cohort of vedolizumab users, 12% had a concomitant prescription for a thiopurine and 5% had a concomitant prescription for methotrexate.

The crude incidence rate (cIR) of an infection-related hospitalization was 5/100 person-years in the anti-TNF-α cohort and 3/100 person-years in the vedolizumab cohort (Figures 1 & 2). After balancing covariates and adjusting for confounders, compared with anti-TNF-α agents, vedolizumab was associated with a significantly decreased risk of an infection-related hospitalization (aHR 0.47, 95%CI: 0.25–0.86 when comparing vedolizumab to anti-TNF-α, Table 2). The most common infections in descending order are septicemia, Clostridium difficile infection, pneumonia and influenza.

Figure 1:

Safety and effectiveness of vedolizumab versus anti-tumor necrosis factor (TNF)-α agent for adults ≥65 years with inflammatory bowel diseases (IBD) in the United States Medicare Claims Database

Steroid prescription in those who did not have baseline steroid use

Figure 2:

Survival curves for infection-related hospitalization in adults ≥65 years with inflammatory bowel diseases (IBD) treated with vedolizumab versus anti-tumor necrosis factor (TNF)-α agent in the United States Medicare Claims Database

Table 2:

Outcomes for adults ≥65 years with inflammatory bowel disease (IBD) initiating vedolizumab or anti-tumor necrosis factor (TNF)-α in the United States Medicare Claims Database

Outcome for Older Adults Initiating Vedolizumab  (Reference: older adults initiating Anti-TNF)	SMRW aHR for all IBD (95% CI)	SMRW aHR for CD (95% CI)	SMRW aHR for UC (95% CI)
Infection-related Hospitalization	0.33 (0.15–0.72)	0.33 (0.13–0.85)	0.68 (0.30–1.56)
IBD-related Hospitalization	1.39 (0.87–2.22)	1.36 (0.73–2.52)	1.03 (0.56–1.87)
IBD-related Surgery	0.96 (0.57–1.63)	0.79 (0.37–1.65)	0.91 (0.45–1.86)
Steroid* Prescription Post-Induction	0.93 (0.73–1.18)	0.86 (0.62–1.19)	1.07 (0.74–1.57)
All-Cause Hospitalization	0.92 (0.74–1.14)	0.98 (0.76–1.26)	0.85 (0.62–1.15)

SMRW aHR: Standardized Morbidity Ratio Weighted adjusted Hazard Ratio

CD: Crohn’s disease

UC: Ulcerative colitis

^*Corticosteroid prescription in those who did not have baseline steroid prescription

The cIR of IBD-related hospitalizations was 6/100 person-years with both anti-TNF-α and vedolizumab (aHR 1.20, 95%CI: 0.78–1.84). The cIR for IBD-related surgery was 5/100 person-years with anti-TNF-α’s and 4/100 person-years with vedolizumab. There was also no significant difference in IBD-related surgery (aHR 0.87, 95%CI: 0.53–1.43). Older adults starting an anti-TNF agent had a cIR of 36/100 person-years for needing a new corticosteroid prescription after biologic initiation compared with 38/100 person-years for older adults initiating vedolizumab. However, after weighting and adjusting, there was no significant difference (aHR 0.93, 95%CI: 0.73–1.18). All-cause hospitalizations were also not significantly different between the two cohorts: (aHR 0.92, 95%CI: 0.76–1.11). An as-treated analysis and an analysis using all available look back data also revealed similar findings. A sensitivity analysis excluding those with concomitant immunomodulator use observed a similar, significantly lower, risk of infection-related hospitalizations for vedolizumab new-users without a significant difference in any of the effectiveness analyses.

We performed 3 stratified analyses by CCI score (0 and ≥1), excluding those who had concomitant immunomodulator use and type of IBD. Vedolizumab initiators with a CCI score of 0 had a significantly lower risk for IBD-related surgery compared with anti-TNF initiators without significant co-morbidities (aHR 0.19, 95%CI: 0.04–0.89). There were no other differences estimating effectiveness in the analysis stratified by CCI score. Even after excluding those who had concomitant immunomodulator use, vedolizumab users had a significantly lower risk for serious infections (aHR 0.35, 95%CI: 0.16–0.76). When stratified by IBD type, the point estimates paralleled the findings of the main analysis: vedolizumab users had a lower risk for infectionrelated hospitalizations (Table 2).

DISCUSSION

In this comparative safety and effectiveness study of biologic agents used to treat IBD in older adults, we present data from a cohort of IBD patients with a median age of 71 years. This cohort is unique in its inclusion of a substantial number of adults of older ages that are not traditionally represented in studies of IBD – 64% of the study cohort was ≥70 years and 11% were ≥80 years. After weighting the cohorts to account for confounding and adjusting for clinically pertinent covariates, including frailty, we found a significantly decreased risk for infection-related hospitalization in older adults treated with vedolizumab, particularly in those with serious non-IBD comorbidities, but no difference in effectiveness of anti-TNF-α and vedolizumab in older adults. With the rapidly growing population of older adults requiring biologic agents for the management of IBD and the increasing number of biologic agents available, these data from a large and generalizable cohort may provide additional evidence to counsel older patients about IBD treatment options.

Older adults are a vulnerable population. Age-related decreases in physiologic reserve and pharmacodynamic changes suggest that extrapolation of risks and benefits of immunosuppressive therapies from younger adults may not be valid.^{27, 28} Our study extends the work of Adar et al who demonstrated similar effectiveness between anti-TNF-α agents and vedolizumab.¹⁴ In contrast with our study, Adar et al did not note a difference in infections, which. may be due to their inclusion of all infections; it is possible that the risk for serious infections requiring hospitalization may be higher with anti-TNF-α agents. Another explanation for the differing results is that our cohort was older, which increases infectious risks.²⁹

A U.S. claims-based study of IBD patients of all ages initiating anti-TNF agents and vedolizumab demonstrated that vedolizumab was associated with a significantly lower risk of infections requiring hospitalization only in patients with UC.³⁰ While the conclusions may be different due to differences in sample size and inclusion criteria, the incidence rates for serious infections were in the same range as our study (~5/100 p-y), suggesting an overall low risk of serious infections. In another pharmacoepidemiologic study assessing the risk of serious infections in all adult IBD patients, there was no difference between vedolizumab and anti-TNF users.³¹ Our safety analysis excluded infections such as intra-abdominal and perianal abscesses that may be attributable to disease and not an adverse medication outcome; this may be another potential explanation for the differing conclusions. Interestingly, immunomodulator use, both at baseline and concomitant to biologic initiation, was higher in our cohort than in these two other studies. However, even after excluding patients with concomitant immunomodulator use, our findings were consistent. Contrasting the findings of our study to these large and robust studies highlights the importance of studies focused on older adults, who have different risks for adverse outcomes, especially infections.³² Large and robust pharmacoepidemiologic studies of older adults with IBD are also needed because older adults are not adequately represented in IBD clinical trials.³³ Additionally, our study also accounted for frailty, a construct which approximates vulnerability better than chronologic age alone.³⁴ Frailty has not previously been included in pharmacoepidemiologic studies of IBD and may provide additional explanation for the differential results.

It is increasingly recognized that biological age, as opposed to chronologic age, confers additional vulnerability to infection.³⁵ We previously demonstrated that frailty, a construct that may better approximate biologic age, is a significant risk factor for infections after immunosuppression for the treatment of IBD.²¹ Another study suggested that frailty is associated with an increased risk of serious infections in patients of all ages treated with vedolizumab, but not anti-TNF-α agents.³⁶ In the present comparative effectiveness and safety study, we adjusted for frailty using a validated algorithm. Adjusting for frailty in our analysis had the potential to control for confounding by indication that is inherent in retrospective comparative effectiveness and safety studies.

VARSITY is the only randomized controlled trial to study comparative efficacy and safety of vedolizumab compared with adalimumab for the treatment of UC. However, VARSITY specified an upper age limit as part of the exclusion criteria and participants had a mean age of 40 years, indicating a young population. Additionally, the authors did not specify how many older adults were in the study nor did they present an age-specific sub-group analysis.³⁷ These aspects limit the extrapolation of the results to older adults. While the VARSITY trial was not powered to detect a difference in serious infections, similarly to our study, they reported a lower risk for serious infections in the vedolizumab arm than adalimumab arm.

Our study has a number of strengths. The large sample size, over 1,500 older adults with IBD newly treated with biologics, and a median age of 71 years from a national cohort generates results that are more generalizable to older adults increasingly seen in clinical practice. The robust methodology accounts for numerous confounders and biases. The new-user design addresses immortal time bias. Comparison of biologic agents helps mitigate confounding by indication. Additionally, this is the first comparative effectiveness and safety study of biologic agents in older adults to account for frailty, which may also reduce confounding by indication.

We acknowledge the limitations of our study associated with retrospective, claims-based studies, including the potential for misclassification bias. However, when feasible we utilized validated algorithms for administrative data to minimize this bias. We did not have access to patient-level data, endoscopic evaluations, laboratory values and disease severity that may influence the interpretation of effectiveness. As with all retrospective studies there are likely to be residual unmeasured confounders. Lastly, due to Medicare policies during the study period, despite ensuring a 12-month look back period, it is possible that vedolizumab users had remote anti-TNF exposure.

In summary, in this Medicare-administrative claims cohort of over 1,500 older adults with IBD newly treated with a biologic agent, we found that older IBD patients treated with vedolizumab had a lower risk of infections requiring hospitalization. We found no significant difference in outcomes approximating effectiveness between anti-TNF-α agents and vedolizumab. While this study provides real world evidence on the influence of age on risks for patients, it is also important to recognize the importance of tailoring treatments to the patient and disease severity. As the number of treatment options for IBD expand and clinical trials do not proportionately represent the rapidly growing population of older adults with IBD, additional large comparative safety and effectiveness studies are needed in older adults with IBD.

What you need to know.

Background:

Despite the rising numbers of older adults with inflammatory bowel diseases (IBD), there are few studies regarding safety and effectiveness of IBD treatments in older adults

Findings:

Adults ≥65 years with IBD newly initiating vedolizumab have a lower risk for an infection-related hospitalization than older adults newly initiating anti-TNF agents. No significant differences were found in IBD-related hospitalizations, IBD-related surgery and new corticosteroid use.

Implications for Patient Care:

These data can inform decision making and help counsel older adults with IBD newly initiating biologic agents