Summary
Background
Direct comparisons are lacking between vedolizumab and tumour necrosis factor (TNF)‐antagonist therapy in Crohn's disease (CD).
Aim
To compare safety and effectiveness of vedolizumab and TNF‐antagonist therapy in adult CD patients.
Methods
Retrospective observational cohort (May 2014–December 2017) propensity score‐weighted comparison of vedolizumab vs TNF‐antagonist therapy (infliximab, adalimumab, certolizumab) in CD. Propensity scores were weighted for age, prior treatments, disease complications, extent and severity, steroid dependence, and concomitant immunosuppressive drug use. The primary outcome was comparative risk for infections or non‐infectious serious adverse events (requiring antibiotics, antivirals, antifungals, hospitalisation, or treatment discontinuation, or resulting in death). Secondary comparative effectiveness outcomes were clinical remission (resolution of CD‐related symptoms), steroid‐free clinical remission and endoscopic remission (absence of ulcers/erosions).
Results
We included 1266 patients (n = 659 vedolizumab). Rates of non‐infectious serious adverse events (odds ratio [OR] 0.072, 95% confidence interval [CI] 0.012‐0.242), but not serious infections (OR 1.183, 95% CI 0.786‐1.795), were significantly lower with vedolizumab vs TNF‐antagonist therapy. Safety comparisons for non‐infectious serious adverse events remained significant after adjusting for differences in duration of exposure. No significant difference was observed between vedolizumab and TNF‐antagonist therapy for clinical remission (hazard ratio [HR] 0.932, 95% CI 0.707‐1.228), steroid‐free clinical remission (HR 1.250, 95% CI 0.677‐2.310) or endoscopic remission (HR 0.827, 95% CI 0.595‐1.151). TNF‐antagonist therapy was associated with higher treatment persistence compared with vedolizumab.
Conclusions
There was a lower risk of non‐infectious serious adverse events, but not serious infections, with vedolizumab vs TNF‐antagonist therapy, with no significant difference for achieving disease remission.
1. INTRODUCTION
Crohn's disease (CD) is an inflammatory bowel disease that can affect the entire intestinal tract and is characterised by mucosal ulcerations, diarrhoea, and abdominal pain. Over time a subset of patients can progress to development of stricturing or penetrating disease complications. Treatment with biologics has been shown to prevent hospitalisation and progression to surgery, with tumour necrosis factor (TNF) antagonists representing the mainstay of biologic therapy. 1 Although effective, TNF antagonists can be associated with serious and sometimes life‐threatening treatment‐related adverse events, including serious or opportunistic infections, and malignancy. 2 Alternative biologic therapies with more targeted mechanisms of action would potentially be advantageous to help avoid these off‐target risks of TNF antagonists.
Vedolizumab, a monoclonal anti‐integrin antibody that targets the α4β7 integrin receptor, is approved for the treatment of moderately to severely active CD. The integrated safety analysis of the phase 3 clinical trial programmes and long‐term safety extension studies observed no increased risk for serious infections with vedolizumab compared to placebo, 3 and our own clinical practice experiences have similarly observed vedolizumab to be well tolerated, with a low rate of serious infections or serious adverse events. 4 The mechanism of action, combined with its observed safety profile in routine practice, has led to the consideration that vedolizumab is potentially safer than TNF‐antagonist therapy in CD. However, indirect comparisons of phase 3 clinical trials have suggested that vedolizumab is the least effective treatment option for achieving disease remission in CD, 5 and no high‐quality head‐to‐head comparisons have been done to specifically assess the theoretical comparative safety advantage.
We therefore studied the comparative safety and effectiveness of vedolizumab and TNF‐antagonist therapy in adult CD patients using a multicentre propensity score (PS)–weighted cohort study. Using patient‐level data from medical records, we compared the relative safety for developing serious infections or serious adverse events, as well as the comparative effectiveness assessment for achievement of disease remission.
2. METHODS
We followed Good ReseArch for Comparative Effectiveness (GRACE) principles and good practice recommendations from the joint International Society for Pharmacoeconomics and Outcomes Research and the International Society for Pharmacoepidemiology for real‐world data comparative effectiveness studies. 6 , 7 The study protocol was posted to the Health Services Research Project website (https://wwwcf.nlm.nih.gov/hsr_project/home_proj.cfm), and the results are reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for cohort studies and reporting guidelines for PS analyses. 8 , 9
2.1. Study design and data source
This is a retrospective review of a North American‐based consortium registry. 10 In brief, this is a multicentre collaborative research group in which outcomes are pooled for consecutive CD patients treated with biologics. Institutional review board approval was obtained from each site for ongoing data collection and transfer. Data were collected individually by sites using a standardised data collection form and transferred (after de‐identification) to the coordinating site (University of California, San Diego) for data compilation and analysis. The current analysis represents data collected between May 2014 and December 2017.
2.2. Participants
Patients from the consortium registry were included in the current analysis if they had (a) a confirmed diagnosis of CD based on clinical, endoscopic or histologic data; (b) active clinical symptoms attributed to CD prior to biologic therapy; and (c) at least one clinical or endoscopic follow‐up after biologic initiation irrespective of response status after induction.
2.3. Variables
Data on variables of interest were collected from sites, including patient characteristics (age at diagnosis, age at biologic initiation, gender, smoking status, body mass index [BMI]), disease characteristics (prior hospitalisations, disease‐related complications, extraintestinal manifestations, and phenotype classified according to Montreal sub‐classifications), and treatment history (steroids, immunomodulators and TNFα antagonists; duration of use; indication for discontinuation; and complications). Variables of interest specific to biologic agent use were baseline disease severity or activity (endoscopic, radiographic, biochemical or clinical assessments), concomitant treatments (steroids or immunomodulators: azathioprine, 6‐mercaptopurine, methotrexate) and follow‐up assessments (endoscopic, radiographic or clinical assessments).
2.4. Outcomes
Comparative safety outcomes were serious infections and serious adverse events. Serious infection was defined as any infection occurring after biologic initiation that required antibiotics, antifungals or antivirals or resulted in discontinuation of biologic therapy, hospitalisation or death. Serious adverse events were defined as having any infectious or noninfectious adverse event after biologic initiation that required antibiotics, antifungals, or antivirals or resulted in discontinuation of biologic therapy, hospitalisation or death. 11 Post‐hoc comparisons were made for noninfectious serious adverse events because of the observed differences in event rates for this outcome in our population, and for the safety outcome of serious infections, limited to vedolizumab and TNF‐antagonist monotherapy subgroups, because of the known risk of infections with concomitant immunosuppressive use.
Comparative effectiveness outcomes were clinical remission, steroid‐free clinical remission, endoscopic remission and treatment persistence. Clinical remission was defined by complete resolution of CD‐related symptoms based on the physician global assessment. Steroid‐free remission was only assessed in those patients on either prednisone or budesonide at the initiation of biologic therapy and was defined as achieving clinical remission, tapering off steroids, and the absence of a subsequent steroid prescription within 1 month of achieving remission. Endoscopic remission was defined as absence of ulcers or erosions, and was limited to patients with confirmed ulcerations and/or erosions at baseline prior to treatment initiation. Treatment persistence took into consideration the occurrence of surgery, and patients were censored at the time of surgery irrespective of whether they continued therapy post‐operatively because this was considered a treatment failure. The coordinating site used de‐identified endoscopy reports to confirm endoscopic remission status, and any discrepancies were resolved through consensus between the study sites and the coordinating site.
2.5. Statistical analyses
Because of the nonrandomised nature of this study, inverse probability weighting (IPW) with PS was used to estimate the average treatment effect (ATE) of vedolizumab vs TNF antagonist in our population. The PS provides an estimate of the individual probability of being treated with the intervention based on information available for that individual, to help account for treatment selection bias inherent in routine practice. 12 , 13 , 14
For effectiveness outcomes, a Cox proportional cause‐specific hazards model was used, with time to surgery treated as a competing risk for effectiveness. 15 , 16 For the safety outcomes, a logistic regression model was used. Sensitivity analyses were done using optimal full match to assess for consistency in estimates across methodologies. 13 , 17 , 18 , 19 Safety data were further reported based on patient years of exposure and after limiting the cohort to patients with at least 12 months of follow‐up.
2.5.1. Propensity score model
Propensity score were calculated using R package “twang” (Toolkit for Weighting and Analysis of Nonequivalent Groups), 20 which estimates PS using boosted regression as the predicted probability of starting treatment with vedolizumab vs TNF‐antagonist therapy, conditioned on the measured baseline variables thought to be confounders or predictors for the outcome of interest. Investigators used a combination of prior published literature, clinical experience and data availability within the current data set to generate a list of potential prognostic variables for consideration. An investigator‐driven approach for confounding evaluation was chosen based on causal knowledge. 21 In addition, care was taken not to include those variables that were strongly correlated with exposure but only weakly correlated with the outcome. 22 , 23 , 24
The final set of variables included for the clinical remission PS model was prior TNF‐antagonist exposure and number of prior TNF antagonists to which the patient was exposed, disease extent (isolated small bowel, ileocolonic, isolated colonic), history of fistulising disease, prior bowel surgery, disease phenotype (stricturing/penetrating or not), clinical disease severity (severe vs nonsevere based on the physician global assessment), CD‐related hospitalisation within the preceding 1 year, baseline steroid dependency or refractoriness, concomitant steroid use or concomitant immunomodulator use. Baseline C‐reactive protein (CRP), albumin and BMI were not included because >25% of the cohort had missing baseline data for these variables. Post‐hoc assessment of these variables revealed that they had no prognostic significance for the primary safety outcomes (serious infection or serious adverse events) or the primary effectiveness outcome (clinical remission), and therefore were unlikely to create an unmeasured confounder bias on the comparative estimates.
Stabilised weights were obtained and further trimmed to be within (0.1, 10), if necessary, before they were used for IPW approaches. 14 Adequacy of the PS model was examined by plotting the PS distributions in the vedolizumab vs TNF‐antagonist groups and by the standardised mean difference of each covariate before and after weighting. The PS model was fit separately for all comparisons of effectiveness and safety between vedolizumab and TNF antagonists 25 and for subgroup analyses. 25 , 26
2.6. Subgroup analyses
A priori subgroup analyses were specified for the comparison of vedolizumab to TNF‐antagonist therapy in TNF‐antagonist–naïve and –exposed patients and for vedolizumab vs infliximab and vedolizumab vs subcutaneous TNF antagonists (adalimumab or golimumab) separately. Subgroup comparison in TNF‐antagonist–naïve and –exposed groups was conducted given prior evidence supporting an increased risk of serious infections and reduction in effectiveness with vedolizumab in TNF‐antagonist–exposed individuals. 3 , 27 , 28 Subgroup comparisons to infliximab and subcutaneous TNF antagonists were performed separately to assess for the potential influence of medical (vedolizumab or infliximab) or pharmacy (subcutaneous TNF antagonists) benefits and market access as a determinant of treatment choice and comparative estimates. 29
Exploratory post‐hoc subgroup analyses were performed for the comparison of vedolizumab to TNF‐antagonist therapy stratified by disease location and disease duration. Disease duration has been observed to significantly impact the achievement of disease remission with vedolizumab, 30 and ileal disease location has been observed to significantly impact the achievement of disease remission across all biologics. 31 To compare whether the hazard ratio (HR) of vedolizumab vs TNF‐antagonist therapy was significantly different across disease location and disease duration subgroups, we performed the Wald test with 2 degrees of freedom in the PS‐weighted analysis.
Secondary post‐hoc analyses were also performed to understand if observed differences in duration of treatment exposure had an effect on observed differences in adverse events and comparisons between vedolizumab and TNF‐antagonist therapy.
2.7. Power calculation
In this observational nonrandomised cohort study, we used the observed event rates in the control group (TNF‐antagonist–treated CD patients) to determine power calculations for unweighted comparisons of safety and effectiveness outcomes using the methodology described by Cohen et al. 1 , 32 For serious infections and serious adverse events we would need an observed event rate of 4% and 7% respectively in the vedolizumab treatment group to be powered for the comparison. For clinical remission we would need 256 observed events to achieve 90% power if the HR was 1.5 or higher between the two groups. If the HR of the effectiveness event were 2.0 or higher, we would need 88 events to achieve 90% power.
3. RESULTS
3.1. Baseline demographics
A total of 1266 patients were included, of whom 659 were treated with vedolizumab, 305 with infliximab and 302 with subcutaneous TNF‐antagonist agents. Baseline demographics are reported in Table 1. The median duration of exposure was higher in the TNF‐antagonist group (508 days, interquartile range [IQR] 311‐786; 932 patient years) vs the vedolizumab group (316 days, IQR 194‐454; 603 patient years). The median duration of treatment persistence without the need for surgery was higher in the TNF‐antagonist group (539 days, IQR 324‐818) vs the vedolizumab group (340 days, IQR 213‐483). At 6 and 12 months, after accounting for the occurrence of surgery as a censoring event, a larger number of TNF‐antagonist–treated patients remained (531 and 417) compared with vedolizumab‐treated patients (506 and 261). (Figure 1).
TABLE 1.
Baseline patient characteristics
| Infliximab (n = 305) | Subcutaneous TNF antagonists (n = 302) | Vedolizumab (n = 659) | |
|---|---|---|---|
| Patient years of exposure | 477 | 458 | 602 |
| Age (diagnosis), mean years (SD) | 36.34 (15.11) | 39.07 (16.21) | 39.80 (15.39) |
| Age (biologic initiation), mean years (SD) | 28.99 (14.53) | 27.80 (13.76) | 25.59 (13.81) |
| Gender (female), n (%) | 142 (46.6) | 160 (53.0) | 381 (57.8) |
| BMI, mean kg/m2 (SD) | 25.19 (6.00) | 25.25 (5.89) | 25.48 (6.87) |
| Smoking status, n (%) | |||
| Current | 28 (9.2) | 36 (11.9) | 69 (10.5) |
| Former | 54 (17.7) | 58 (19.2) | 112 (17.0) |
| Never | 223 (73.1) | 208 (68.9) | 478 (72.5) |
| Disease duration, mean years (SD) | 3 (10) | 6 (17) | 12 (13) |
| Ever hospitalised? n (%) | |||
| Never | 119 (39.0) | 128 (42.4) | 139 (21.1) |
| Yes (in the last year) | 119 (39.0) | 60 (19.9) | 240 (36.4) |
| Yes (not in the last year) | 67 (22.0) | 114 (37.7) | 280 (42.5) |
| CRP, mean mg/L (SD) | 1.4 (6.6) | 0.7 (1.9) | 4.6 (15.83) |
| Albumin, mean g/dL (SD) | 3.87 (0.60) | 4.09 (0.50) | 3.87 (0.54) |
| Rheumatic EIM (yes), n (%) | 54 (17.7) | 62 (20.5) | 143 (21.7) |
| Ophthalmologic EIM (yes), n (%) | 8 (2.6) | 6 (2.0) | 15 (2.3) |
| Dermatologic EIM (yes), n (%) | 17 (5.6) | 19 (6.3) | 47 (7.1) |
| Hepatic EIM (yes), n (%) | 1 (0.3) | 4 (1.3) | 11 (1.7) |
| Disease extent, n (%) | |||
| L1 | 73 (23.9) | 73 (24.2) | 104 (15.8) |
| L2 | 71 (23.3) | 57 (18.9) | 140 (21.2) |
| L3 | 160 (52.5) | 167 (55.3) | 413 (62.7) |
| L4 | 1 (0.3) | 5 (1.7) | 2 (0.3) |
| Phenotype, n (%) | |||
| B1 | 115 (39.1) | 89 (29.6) | 214 (32.6) |
| B2 | 43 (14.6) | 59 (19.6) | 161 (24.5) |
| B3 | 136 (46.3) | 153 (50.8) | 282 (42.9) |
| Fistulising disease history (yes), n (%) | 138 (47.1) | 111 (37.0) | 240 (36.5) |
| Disease severity, n (%) | |||
| Mild | 30 (9.8) | 48 (15.9) | 101 (15.3) |
| Moderate | 171 (56.1) | 164 (54.3) | 330 (50.1) |
| Severe | 104 (34.1) | 90 (29.8) | 228 (34.6) |
| Endoscopic severity, n (%) | |||
| Mild | 48 (21.0) | 57 (25.9) | 102 (22.3) |
| Moderate | 93 (40.6) | 77 (35.0) | 184 (40.3) |
| Severe | 88 (38.4) | 86 (39.1) | 171 (37.4) |
| Prior surgery (yes), n (%) | 128 (42.0) | 142 (47.0) | 400 (60.7) |
| Concomitant IM (yes), n (%) | 145 (47.5) | 136 (45.0) | 272 (41.3) |
| Concomitant steroid (yes), n (%) | 83 (27.2) | 80 (26.5) | 302 (45.8) |
| Steroid dependency (yes), n (%) | 47 (15.4) | 60 (19.9) | 247 (37.5) |
| Prior TNF antagonist (yes), n (%) | 144 (47.2) | 172 (57.0) | 598 (90.7) |
| Number of prior TNF antagonists, n (%) | |||
| 0 | 161 (52.8) | 130 (43.0) | 60 (9.1) |
| 1 | 116 (38.0) | 130 (43.0) | 161 (24.4) |
| 2 | 25 (8.2) | 36 (11.9) | 253 (38.4) |
| 3 | 3 (1.0) | 5 (1.7) | 183 (27.8) |
| 4 | 0 (0.0) | 1 (0.3) | 2 (0.3) |
Abbreviations: BMI, body mass index; CRP, C‐reactive protein; EIM, extraintestinal manifestation; IM, immunomodulator; SD, standard deviation; TNF, tumour necrosis factor.
FIGURE 1.
Treatment persistence. Kaplan Meier plot for duration of follow‐up. TNF, tumour necrosis factor. Y‐axis is proportion of patients still on therapy and X‐axis is time point of follow‐up
There was a slight female predominance among patients treated with vedolizumab compared with infliximab or subcutaneous TNF antagonist (57.8% vs 53.0% vs 46.6%) and a longer mean disease duration (12 vs 6 vs 3 years). Baseline mean CRP was higher in vedolizumab‐treated patients (4.6 vs 0.7 vs 1.4 mg/L), and vedolizumab‐treated patients more often had prior bowel resection (60.7% vs 47.0% vs 42.0%). Only 9.1% of vedolizumab‐treated patients were TNF‐antagonist naïve compared with 43.0% of subcutaneous TNF‐antagonist–treated patients and 52.8% of infliximab‐treated patients. Vedolizumab‐treated patients were more often steroid dependent at baseline (37.5% vs 15.4% vs 19.9%) and more often taking concomitant steroids (45.8% vs 27.2% vs 26.5%). Prior TNF‐antagonist exposure and number of prior TNF antagonists used constituted the most imbalanced variables at baseline between groups, but after weighting these were both well balanced, along with all other variables as assessed by standardised mean difference (Figure 2).
FIGURE 2.
Standardised mean difference of variables included before and after weighting in the full sample. Age (Biol.), age at initiation of treatment with biologic agent; ATE, average treatment effect; IM, immunomodulator; N.Anti. TNF exposed, number of anti‐TNF agents to which patient was exposed; TNF, tumour necrosis factor. For SMD plots: SMD before (red) and after (blue) weighting, low absolute value of SMD (usually between 0 and 0.1) means good balance of the distribution of a variable between the treatment and the placebo. For PS distribution plots: Propensity score means: the probability of receiving treatment given the covariates
3.2. Safety outcomes
3.2.1. Serious infection rates and events
A total of 47 vedolizumab‐treated and 47 TNF‐antagonist–treated patients developed serious infections, with one death in each group due to serious infections. This equates to an unadjusted rate of 7.8 per 100 patient years for vedolizumab‐treated patients and 5.1 per 100 patient years for the TNF‐antagonist treated patients. When limiting the cohort to patients with at least 12 months of follow‐up, the rate of serious infections was comparable between vedolizumab‐treated (n = 24; 9.2%, 6.5 per 100 patient years) and TNF‐antagonist treated (n = 39; 9.7%, 4.2 per 100 patient years) patients.
The most common serious infections were abscess (n = 8/47 vedolizumab, n = 6/47 TNF antagonist), Clostridium difficile (n = 7/47 vedolizumab, n = 5/47 TNF antagonist), other enteric infections (n = 5/47 vedolizumab, n = 7/47 TNF antagonist), skin infections (n = 3/47 vedolizumab, n = 2/47 TNF antagonist), upper and lower respiratory tract infections (n = 5/47 vedolizumab, n = 5/47 TNF antagonist) and shingles (n = 2/47 vedolizumab, n = 2/47 TNF antagonist). Two vedolizumab‐treated patients developed gram‐negative rod bacteraemia, one of whom died. One TNF‐antagonist–treated patient died of Gram‐negative rod bacteraemia.
3.2.2. Non‐infectious serious adverse event rates and events
Three vedolizumab‐treated and 25 TNF‐antagonist–treated patients developed non‐infectious serious adverse events. This equates to an unadjusted rate of 0.5 per 100 patient years for vedolizumab‐treated patients and 2.7 per 100 patient years for the TNF‐antagonist treated patients. When limiting the cohort to patients with at least 12 months of follow‐up, the rate of non‐infectious serious adverse events was lower in vedolizumab‐treated patients (n = 2; 0.8%, 0.54 per 100 patient years) as compared to TNF‐antagonist–treated patients (n = 14; 3.5%, 1.7 per 100 patient years).
For vedolizumab‐treated patients, these events were severe arthralgias requiring therapy discontinuation (n = 3). For TNF‐antagonist–treated patients, events included hypersensitivity or infusion reactions (n = 6), drug‐induced psoriasis (n = 6), drug‐induced lupus (n = 5), severe liver function test abnormalities (n = 3), skin rash (n = 2), lung cancer (n = 1) and jaw or hip necrosis (n = 2).
3.2.3. Comparative safety
We observed no significant difference in risk of serious infections (OR 1.183, 95% CI 0.786‐1.795) or serious adverse events (OR 0.751, 95% CI 0.519‐1.086) between vedolizumab‐ and TNF‐antagonist–treated patients (Table 2). The difference in risk of noninfectious serious adverse events specifically was significantly lower with vedolizumab vs TNF‐antagonist therapy (IPW ATE: OR 0.072, 95% CI 0.012‐0.242; full match: OR 0.150, 95% CI 0.035‐0.441). During post‐hoc analyses, we observed no significant difference in risk of serious infection between vedolizumab monotherapy and TNF‐antagonist monotherapy (IPW ATE: OR 0.759, 95% CI 0.300‐1.876; full match: OR 1.426, 95% CI 0.563‐3.782).
TABLE 2.
Comparative safety of vedolizumab to TNF‐antagonist therapy in Crohn's disease
| Overall cohort | TNF‐antagonist naive | TNF‐antagonist exposed | |
|---|---|---|---|
| Serious infection | |||
| Unweighted | 0.887 (0.581‐1.352) | 1.526 (0.579‐3.594) | 0.809 (0.484‐1.376) |
| IPW ATE | 1.183 (0.786‐1.795) | 1.856 (0.681‐4.507) | 1.111 (0.657‐1.944) |
| Full match | 0.970 (0.638‐1.474) | 2.130 (0.587‐9.232) | 1.484 (0.905‐2.466) |
| Serious adverse event | |||
| Unweighted | 0.591 (0.403‐0.862) | 1.027 (0.424‐2.236) | 0.571 (0.357‐0.915) |
| IPW ATE | 0.751 (0.519‐1.086) | 1.082 (0.414‐2.478) | 0.702 (0.441‐1.133) |
| Full match | 0.887 (0.581‐1.352) | 1.526 (0.579‐3.594) | 0.809 (0.484‐1.376) |
Values are odds ratio (95% confidence interval).
Abbreviations: IPW ATE, inverse probability weighting average treatment effect; TNF, tumour necrosis factor.
Post‐hoc analyses for duration of treatment exposure observed no significant association between duration of treatment exposure and risk of serious adverse events (OR 1.01, 95% CI 0.99‐1.03, P = 0.44), serious infections (OR 1.01, 95% CI 0.99‐1.03, P = 0.39), or noninfectious serious adverse events (OR 1.00, 95% CI 0.96‐1.04, P = 0.96). After forcing treatment exposure in as an effect modifier for comparisons between vedolizumab and TNF‐antagonist therapy, we again observed that vedolizumab‐treated CD patients were significantly less likely to develop noninfectious serious adverse events relative to TNF‐antagonist‒treated CD patients (IPW ATE: OR 0.940, 95% CI 0.892‐0.984, P = 0.013; IPW ATT: OR 0.915, 95% CI 0.862‐0.966, P = 0.002; IPW Full Match: OR 0.862, 95% CI 0.802‐0.920, P < 0.001).
3.3. Comparative effectiveness
We observed no significant difference for achieving clinical remission (HR 0.932, 95% CI 0.707‐1.228), steroid‐free clinical remission (HR 1.250, 95% CI 0.677‐2.310), or endoscopic remission (HR 0.827, 95% CI 0.595‐1.151) between vedolizumab and TNF‐antagonist–treated patients. Results were consistent when comparing vedolizumab to infliximab and subcutaneous TNF‐antagonist agents separately in the entire cohort (Table 3).
TABLE 3.
Comparative effectiveness of vedolizumab to TNF‐antagonist therapy in Crohn's disease
| Overall | VDZ vs IFX | VDZ vs SQ TNF antagonist | |
|---|---|---|---|
| Clinical remission | |||
| Unweighted | 0.812 (0.671‐0.982) | 0.697 (0.557‐0.872) | 0.940 (0.743‐1.189) |
| IPW ATE | 0.932 (0.707‐1.228) | 0.692 (0.472‐1.014) | 1.022 (0.714‐1.404) |
| Full match | 0.917 (0.470‐1.791) | 0.737 (0.301‐1.806) | 0.926 (0.415‐2.066) |
| Steroid‐free clinical remission a | |||
| Unweighted | 0.953 (0.628‐1.446) | 0.635 (0.398‐1.012) | 1.647 (0.861‐3.154) |
| IPW ATE | 1.250 (0.677‐2.310) | 0.695 (0.295‐1.641) | 1.717 (0.665‐4.432) |
| Full match | 1.262 (0.416‐3.828) | 0.312 (0.126‐0.775) | 2.365 (0.540‐10.358) |
| Endoscopic remission | |||
| Unweighted | 0.727 (0.575‐0.918) | 0.607 (0.465‐0.793) | 0.875 (0.660‐1.160) |
| IPW ATE | 0.827 (0.595‐1.151) | 0.696 (0.450‐1.076) | 1.026 (0.660‐1.547) |
| Full match | 1.307 (0.614‐2.780) | 1.269 (0.472‐3.411) | 1.159 (0.477‐2.815) |
Values are hazard ratio (95% confidence interval).
Abbreviations: IFX, infliximab; IPW ATE, inverse probability weighting average treatment effect; SQ, subcutaneous; TNF, tumour necrosis factor; VDZ, vedolizumab.
Steroid‐free clinical remission limited to patients taking concomitant steroids at baseline. Endoscopic remission limited to patients with follow‐up assessment of endoscopic disease activity (n = 424 TNF‐antagonist; n = 413 vedolizumab).
Among TNF‐antagonist–naïve patients, we observed a significant difference for the achievement of clinical remission (HR 1.861, 95% CI 1.059‐3.272) and steroid‐free clinical remission (HR 5.608, 95% CI 1.471‐21.374) that favoured vedolizumab over subcutaneous TNF‐antagonist agents. This comparison was not significantly different between vedolizumab and infliximab in TNF‐antagonist–naïve patients (Table 4).
TABLE 4.
Comparative effectiveness of vedolizumab to TNF‐antagonist therapy in Crohn's disease stratified by TNF‐antagonist exposure
| TNF‐antagonist naive | TNF‐antagonist exposed | |||||
|---|---|---|---|---|---|---|
| Overall | VDZ vs IFX | VDZ vs SQ TNF antagonist | Overall | VDZ vs IFX | VDZ vs SQ TNF antagonist | |
| Clinical remission | ||||||
| Unweighted | 1.520 (0.979‐2.362) | 1.413 (0.879‐2.271) | 1.660 (1.014‐2.719) | 0.763 (0.604‐0.964) | 0.582 (0.435‐0.779) | 0.960 (0.713‐1.292) |
| IPW ATE | 1.654 (1.029‐2.659) | 1.361 (0.789‐2.347) | 1.861 (1.059‐3.272) | 0.821 (0.588‐1.146) | 0.557 (0.331‐0.937) | 0.969 (0.612‐1.536) |
| Full match | 0.923 (0.459‐1.857) | 1.087 (0.544‐2.173) | 1.007 (0.392‐2.590) | 0.972 (0.520‐1.817) | 0.362 (0.169‐0.777) | 1.266 (0.485‐3.306) |
| Steroid‐free clinical remission a | ||||||
| Unweighted | 1.858 (0.804‐4.295) | 1.210 (0.504‐2.907) | 4.637 (1.236‐17.390) | 0.959 (0.558‐1.646) | 0.672 (0.349‐1.292) | 1.394 (0.631‐3.080) |
| IPW ATE | 2.002 (0.763‐5.255) | 1.156 (0.504‐3.309) | 5.608 (1.471‐21.374) | 1.014 (0.469‐2.192) | 0.452 (0.145‐1.408) | 1.398 (0.424‐4.611) |
| Full match | 2.530 (0.971‐6.596) | 0.930 (0.351‐2.463) | 11.127 (1.798‐68.854) | 1.569 (0.374‐6.577) | 0.319 (0.117‐0.871) | 6.170 (1.574‐24.181) |
| Endoscopic remission | ||||||
| Unweighted | 0.691 (0.320‐1.492) | 0.663 (0.301‐1.462) | 0.728 (0.325‐1.632) | 0.823 (0.622‐1.090) | 0.600 (0.424‐0.849) | 1.033 (0.730‐1.460) |
| IPW ATE | 0.900 (0.366‐2.213) | 0.771 (0.323‐1.840) | 0.907 (0.364‐2.260) | 0.864 (0.573‐1.301) | 0.774 (0.405‐1.478) | 1.091 (0.640‐1.859) |
| Full match | 0.864 (0.321‐2.328) | 0.418 (0.181‐0.965)b | 1.851 (0.623‐5.493) | 1.038 (0.427‐2.522) | 1.278 (0.354‐4.616) | 1.271 (0.486‐3.325) |
Values are hazard ratio (95% confidence interval).
Abbreviations: IFX, infliximab; IPW ATE, inverse probability weighting average treatment effect; SQ, subcutaneous; TNF, tumour necrosis factor; VDZ, vedolizumab.
Steroid‐free clinical remission limited to patients taking concomitant steroids at baseline. Endoscopic remission limited to patients with follow‐up assessment of endoscopic disease activity (n = 424 TNF‐antagonist; n = 413 vedolizumab).
Among TNF‐antagonist–exposed patients, point estimates for the comparison of vedolizumab and infliximab favoured infliximab for the achievement of clinical remission and steroid‐free remission, and this reached statistical significance using full match (clinical remission: HR 0.362, 95% CI 0.169‐0.777; steroid‐free clinical remission: HR 0.319, 95% CI 0.117‐0.871) (Table 4).
3.4. Impact of disease location and disease duration on effectiveness
Disease location did not significantly impact the comparative effectiveness estimates between vedolizumab and TNF‐antagonist–treated patients (Table S1). Using full match, disease duration significantly impacted the comparative effectiveness estimates between vedolizumab and TNF‐antagonist–treated patients for the outcomes of clinical remission (P = 0.034) and steroid‐free clinical remission (P = 0.018). Point estimates for comparisons within these subgroups suggested that vedolizumab might be favoured over TNF‐antagonist agents in early‐disease CD (≤2 years) and that TNF‐antagonist agents might be favoured over vedolizumab in late‐disease CD (Figure 3; Table S1).
FIGURE 3.
Comparative effectiveness of vedolizumab and TNF‐antagonist therapy stratified by disease duration subgroups. CI, confidence interval; TNF, tumour necrosis factor
4. DISCUSSION
In this routine practice PS‐weighted cohort of over 1200 biologic‐treated CD patients, we observed no significant difference between vedolizumab and TNF‐antagonist therapy for risk of experiencing a serious infection or serious adverse event or for probability of achieving clinical remission, steroid‐free clinical remission or endoscopic remission. Exploratory subgroup analyses suggested that vedolizumab might be superior to subcutaneous TNF‐antagonist therapy for the achievement of clinical remission and steroid‐free clinical remission in TNF‐antagonist–naïve patients, and infliximab might be superior to vedolizumab for the achievement of clinical remission and steroid‐free clinical remission in TNF‐antagonist–exposed patients. During subgroup analyses, we also observed a significant interaction between disease duration and comparative effectiveness estimates, with vedolizumab potentially being favoured in early‐disease CD and TNF antagonists potentially being favoured in later‐disease CD.
We observed no significant difference in treatment‐related serious infection risks between vedolizumab and TNF antagonists, with a comparable event rate being observed in both groups. The risk of serious infections with biologic therapy is largely driven by disease activity and concomitant use of immunosuppressive agents. 4 , 11 , 33 , 34 , 35 The lack of observed difference between vedolizumab and TNF‐antagonist therapy for achievement of disease remission and the higher concomitant use of steroids among the vedolizumab‐treated patients in our cohort may therefore help to explain the lack of observed difference in risk for serious infections between agents. When limiting the serious infection comparison to patients on biologic monotherapy, however, we observed no significant difference in risk of serious infections, which argues against the concomitant immunosuppressive therapy as a determinant of the comparative safety assessment.
The event rate for noninfectious serious adverse events specifically, however, was significantly lower in the vedolizumab‐exposed group (n = 3) than in the TNF‐antagonist–exposed group (n = 25; OR 0.072). This remained significant even after adjusting for differences in treatment exposure between both groups. One of the theoretical advantages of vedolizumab over TNF‐antagonist therapy is the potentially lower rate of noninfectious “off‐target” adverse events due to differences in mechanism of action. Common noninfectious adverse events observed with TNF‐antagonist therapy are hypersensitivity reactions, psoriasis, and lupus, as well as other dermatological or autoimmune‐like disorders. 36 Several of these were observed with TNF‐antagonist therapy in our cohort, whereas none were observed with vedolizumab. Larger population‐based cohort studies are therefore needed to confirm our observations.
We observed no significant difference between vedolizumab and TNF‐antagonist therapy for the achievement of clinical remission, steroid‐free clinical remission, or endoscopic remission, and this was consistent when comparing vedolizumab to infliximab or subcutaneous TNF antagonists separately in the entire cohort or when comparing vedolizumab to TNF‐antagonist agents in TNF‐antagonist–naïve and –exposed subgroups. We did observe TNF‐antagonist therapy to be associated with greater treatment persistence as measured by duration of exposure. It is unclear if this is a function of the time period of observation or if it represents a true difference in treatment persistence between therapies. Prior work from our group has observed that inflammatory bowel disease patients treated with vedolizumab early after its approval and availability in the market were more refractory and had lower persistence rates compared to more recent utilisation. 37 Therefore, it's possible that these differences were related to utilisation patterns in which vedolizumab‐treated patients earlier in the period were simply more refractory requiring early discontinuation and those treated later in the observation period had less time to be exposed to the drug.
Our observational cohort study was not designed to be a noninferiority study, and the safety and effectiveness comparisons were exploratory in nature. Post‐hoc power calculations suggested adequate sample size and event rates for the comparisons, and it is interesting to consider the power of our study to make comparisons understanding the limitations of PS analyses relative to randomised controlled trials, and that in observational effectiveness studies the lack of randomisation degrades the ability to demonstrate noninferiority. A trial between vedolizumab and TNF‐antagonist therapy with an estimated 30% rate for clinical remission, 90% power and 10% noninferiority limit would require 720 participants with equal distribution between groups. A 7.5% noninferiority limit would require 1280 participants, and a 5% noninferiority limit would require 2880 participants. 38 Our cohort of 1266 patients is therefore relatively robust to explore the overall comparative safety and effectiveness of these agents in routine practice.
In subgroup analyses, we observed a significant interaction between disease duration and comparative effectiveness estimates, with vedolizumab potentially being favoured over TNF antagonists in early CD (≤2 years). Disease duration has been observed to significantly impact treatment effectiveness for both vedolizumab and TNF‐antagonist therapy. 30 , 39 , 40 However, a comparison of effectiveness between these two agents in early CD is lacking. Our exploratory subgroup analysis for this important question suggests that perhaps vedolizumab might be more efficacious than TNF‐antagonist therapy in these patients. These results should be interpreted with caution given the relatively small sample size of patients with early CD (≤2 years) in our cohort and the exploratory nature of this comparison. Nonetheless, given the lower risk of noninfectious adverse events with vedolizumab and the incremental risk for noninfectious adverse events with longer duration of use for TNF antagonists, early use of vedolizumab in CD might be an attractive treatment option to optimise disease outcomes while minimising treatment‐related risks. This hypothesis will require formal testing with a clinical trial specifically aimed to assess the comparative effectiveness for achievement of disease remission and reduction in disease and treatment‐related risks between vedolizumab and TNF antagonists to understand relative trade‐offs between efficacy and safety. Furthermore, with the availability of biosimilar agents for TNF antagonists, long‐term maintenance costs will also need to be factored into this strategy.
Our study has several strengths, which include the routine practice nature of our cohort, the robustness of our PS‐weighting methodology, and the relatively modest sample size for comparison estimates. Several important limitations remain. The retrospective observational nature of data carries limitations, lack of standardised numerical clinical disease activity measurements, and the academic centre nature of sites may impact generalisability to routine community practices. The detailed patient data available allowed for a very granular PS‐weighting methodology to be applied that incorporated several markers of disease activity and severity, but residual confounding may still exist which limits the comparison. We also could not adequately account for practice variation with therapeutic drug monitoring or dose escalation during maintenance therapy across sites because of the lack of standardised treatment protocols. Therefore data on dose‐intensification were not captured and represents a limitation, however, it may also represent a strength by allowing for routine practice variation when making comparisons across treatments. Caution should also be taken when interpreting comparisons for secondary outcomes and subgroup analyses given the inherent selection bias across these subgroups, limited sample size and event rate, and lack of significance for the primary comparisons. 41 These secondary analyses such as the impact of disease location and disease duration should be considered hypothesis generating as opposed to hypothesis testing, and should be interested with caution. Finally, the small number of TNF‐antagonist–naïve vedolizumab‐treated patients creates a limited common support for the comparative assessment of vedolizumab to TNF‐antagonist agents in TNF‐antagonist–naïve patients.
In summary, in this large, routine practice, PS‐weighted comparative study, we observed no significant differences between vedolizumab and TNF‐antagonist therapy for risk of serious infections or serious adverse events or for the probability of achieving disease remission. Post‐hoc secondary analyses observed a significant difference between vedolizumab and TNF‐antagonist therapy for the risk of noninfectious adverse events specifically, which remained after accounting for differences in duration of follow‐up, but not for the risk of serious infections between vedolizumab monotherapy and TNF‐antagonist monotherapy. Subgroup analyses also suggested that vedolizumab might be favoured in early‐disease CD, but along these same lines TNF‐antagonist therapy may be favoured in later‐stage CD. Larger well‐powered clinical trials will be needed to confirm the observed impact of disease duration on the comparative effectiveness estimates in our study. In particular, more data are needed on vedolizumab used as a first‐line biologic therapy and as true monotherapy to better understand the risk for infectious adverse events.
AUTHORSHIP
Guarantor of the article: Parambir S. Dulai, MD.
Author contributions: MB and PSD contributed to conceptualisation and planning. RX and PSD contributed to methodology. MB, SV, MF, SK, GT, MR, SA, JM, AW, GK, JLKP, JPC, BB, SS, DF, AW, SC, RH, RU, ES, VP, DH, SC, DL, AS, BES, JFC, SK, EVL, BS, CAS, WJS and PSD contributed to data curation. RX, YZ and PSD contributed to formal statistical analyses. MB, RX, YZ and PSD contributed to manuscript writing. All authors contributed to critical revisions and final approval. PSD contributed to study supervision. All authors approved the final version of the article, including the authorship list.
Supporting information
Table S1
ACKNOWLEDGEMENTS
Declaration of personal interests: Matthew Bohm, Ronghui Xu, Yiran Zhang, Sashidhar Varma, Monika Fischer, Siri Kadire, Gloria Tran, Mahmoud Rahal, Satimai Aniwan, Aaron Weiss, Gursimran Kochhar, James P. Campbell, Adam Winters, and Shreya Chablaney have no potential conflicts of interest to disclose. Joseph Meserve reports travel support from Takeda. Jenna L. Koliani‐Pace reports travel support from Takeda. Brigid Boland receives research support from Takeda and Janssen. Siddharth Singh reports research support from AbbVie, Pfizer; consulting for Pfizer, AbbVie, Takeda, AMAG Pharmaceuticals; career development awards from the American College of Gastroenterology and Crohn's and Colitis Foundation. David Faleck reports travel support from Takeda. Robert Hirten reports speaking or advisory boards for Takeda, Janssen. Ryan Ungaro is a consulting and/or advisory board member for Takeda, Pfizer, and Janssen. Eugenia Shmidt reports travel support from Takeda. Karen Lasch is an employee of Takeda Pharmaceuticals, USA, Inc Vipul Jairaith reports consulting fees from AbbVie, Janssen, Takeda, Sandoz, Ferring, Pfizer, GlaxoSmithKline, Robarts Clinical Trials, Eli Lilly, Arena; speaker fees from: Takeda, Ferring, Janssen, Shire. David Hudesman reports consulting fees from AbbVie, Takeda, Janssen, Pfizer. Shannon Chang reports consulting fees from Oshi. Dana Lukin, reports consulting for AbbVie, Janssen, Salix. Arun Swaminath reports fellowship support from Janssen, AbbVie, Pfizer; grant support from Pfizer. Bruce E. Sands reports consulting/advisory board or honorarium: 4D Pharma, AbbVie, Allergan Sales, Amgen, Arena Pharmaceuticals, Boehringer Ingelheim, Capella BioScience, Celgene, EnGene, Ferring, Gilead, Janssen, Lilly, Lyndra, MedImmune, Oppilan Pharma, Otsuka, Palatin Technologies, Pfizer, Progenity, Rheos Pharmaceuticals, Seres Therapeutics, Synergy Pharmaceuticals, Takeda, Target PharmaSolutions, Theravance Biopharma R&D, Inc, TiGenix, Vivelix Pharmaceuticals, WebMD. Jean‐Frederic Colombel reports consultant or advisory board member for AbbVie, Amgen, Boehringer Ingelheim, Arena Pharmaceuticals, Celgene Corporation, Celltrion, Enterome, Eli Lilly, Ferring Pharmaceuticals, Genentech, Janssen and Janssen, Medimmune, Merck & Co., Nextbiotix, Novartis Pharmaceuticals Corporation, Otsuka Pharmaceutical Development & Commercialization, Inc, Pfizer, Protagonist, Second Genome, Gilead, Seres Therapeutics, Shire, Takeda, Theradiag; speaker for AbbVie, Ferring, Takeda, Celgene Corporation; stock options: Intestinal Biotech Development, Genefit; research grants: AbbVie, Takeda, Janssen and Janssen. Sunanda Kane reports consulting for AbbVie, Merck, Spherix Health, Seres Pharmaceuticals, Samsung Bioepis. Edward V. Loftus reports consulting for Janssen, Takeda, AbbVie, UCB, Amgen, Pfizer, Salix, Mesoblast, Eli Lilly, Celgene, and CVS Caremark; research support from Janssen, Takeda, AbbVie, UCB, Amgen, Pfizer, Genentech, Gilead, Receptos, Celgene, MedImmune, Seres Therapeutics, and Robarts Clinical Trials. Bo Shen reports consulting for Janssen, Salix, AbbVie, Takeda, Theravance, Robarts Clinical Trials. Corey A. Siegel reports consulting/advisory board AbbVie, Amgen, Celgene, Lilly, Janssen, Sandoz, Pfizer, Prometheus, Sebela, Takeda; Speaker AbbVie, Janssen, Pfizer, Takeda; Grant support Crohn's and Colitis Foundation, AHRQ (1R01HS021747‐01), AbbVie, Janssen, Pfizer, Takeda; Intellectual property MiTest Health, LLC, and ColonaryConcepts, LLC; Equity interest MiTest Health and ColonaryConcepts. William J. Sandborn reports personal fees from Kyowa Hakko Kirin, Millennium Pharmaceuticals, Celgene Cellular Therapeutics, Santarus, Salix Pharmaceuticals, Catabasis Pharmaceuticals, Vertex Pharmaceuticals, Warner Chilcott, Cosmo Pharmaceuticals, Ferring Pharmaceuticals, Sigmoid Biotechnologies, Tillotts Pharma, Am Pharma BV, Dr August Wolff, Avaxia Biologics, Zyngenia, Ironwood Pharmaceuticals, Index Pharmaceuticals, Nestle, Lexicon Pharmaceuticals, UCB Pharma, Orexigen, Luitpold Pharmaceuticals, Baxter Healthcare, Ferring Research Institute, Novo Nordisk, Mesoblast Inc, Shire, Ardelyx Inc, Actavis, Seattle Genetics, MedImmune (AstraZeneca), ActoGeniX NV, Lipid Therapeutics GmbH, Eisai, Qu Biologics, Toray Industries Inc, Teva Pharmaceuticals, Eli Lilly, Chiasma, TiGenix, Adheron Therapeutics, Immune Pharmaceuticals, Celgene, Arena Pharmaceuticals; personal fees from Ambrx Inc, Akros Pharma, Vascular Biogenics, Theradiag, Forward Pharma, Regeneron, Galapagos, Seres Health, Ritter Pharmaceuticals, Theravance, Palatin, Biogen, University of Western Ontario (owner of Robarts Clinical Trials); grants and personal fees from Prometheus Laboratories, AbbVie, Gilead Sciences, Boehringer Ingelheim, Amgen, Takeda, Atlantic Pharmaceuticals, Bristol‐Myers Squibb, Genentech, GlaxoSmithKline, Pfizer, Nutrition Science Partners, Receptos, Amgen; grants, personal fees and non‐financial support from Janssen; grants from Broad Foundation, American College of Gastroenterology, Exact Sciences. Parambir S. Dulai is on the steering committee for Takeda; does consulting for Takeda, Janssen; honorarium for speaker events Takeda; travel support Takeda, Janssen; grant support Takeda, Pfizer.
Appendix A. Collaborators: VICTORY Collaboration
Indiana University, Indianapolis, IN, USA: Siri Kadire, MD; Gloria Tran, MD; Mahmoud Rahal, MD. Mayo Clinic, Rochester, MN, USA: Satimai Aniwan, MD. University of California San Diego, La Jolla, CA, USA: Joseph Meserve, MD. Montefiore Medical Center, New York, NY, USA: Aaron Weiss, MD. Dartmouth‐Hitchcock Medical Center, Lebanon, NH: Jenna L. Koliani‐Pace, MD. University of Minnesota, Minneapolis, MN, USA: James P. Campbell, MD. Icahn School of Medicine at Mount Sinai, New York, NY, USA: David Faleck, MD; Adam Winters, MD; Shreya Chablaney, MD.
Appendix B. THE AUTHORS' COMPLETE AFFILIATION LIST
Matthew Bohm, Sashidhar Varma and Monika Fischer: Indiana University, Indianapolis, IN, USA. Ronghui Xu, Yiran Zhang, Brigid Boland, Siddharth Singh, William J. Sandborn and Parambir S. Dulai: University of California San Diego, La Jolla, CA, USA. Gursimran Kochhar and Bo Shen: Cleveland Clinic Foundation, Cleveland, OH, USA. Robert Hirten, Ryan Ungaro, Bruce E. Sands and Jean‐Frederic Colombel: Icahn School of Medicine at Mount Sinai, New York, NY, USA. Eugenia Shmidt: University of Minnesota, Minneapolis, MN, USA. Karen Lasch: Takeda Pharmaceuticals USA, Inc, Deerfield, IL, USA. Vipul Jairaith: University of Western Ontario, London, ON, Canada. David Hudesman and Shannon Chang: New York University, New York, NY, USA. Dana Lukin: Montefiore Medical Center, New York, NY, USA. Arun Swaminath: Lenox Hill Hospital, New York, NY, USA. Sunanda Kane and Edward V. Loftus, Jr: Mayo Clinic, Rochester, MN, USA. Corey A. Siegel: Dartmouth‐Hitchcock Medical Center, Lebanon, NH, USA.
Bohm M, Xu R, Zhang Y, et al; the VICTORY Collaboration . Comparative safety and effectiveness of vedolizumab to tumour necrosis factor antagonist therapy for Crohn's disease. Aliment Pharmacol Ther. 2020;52:669–681. 10.1111/apt.15921
The Handling Editor for this article was Professor Jonathan Rhodes, and it was accepted for publication after full peer‐review.
Matthew Bohm and Ronghui Xu equally contributed as first author.
See Appendix A for list of Collaborators.
The complete list of authors' affiliation are listed in Appendix B.
Funding information
Takeda Pharmaceuticals provided funding for statistical support to analyse the data; however, Takeda Pharmaceuticals and associated employees did not have access to any of the data. Data were collected at each site individually and transferred as de‐identified data to the coordinating centre (University of California San Diego) for compilation and analyses. All data analyses were performed at the University of California San Diego by consortium investigators or statisticians, independent of Takeda Pharmaceuticals.
Contributor Information
Parambir S. Dulai, Email: pdulai@ucsd.edu.
the VICTORY Collaboration:
Siri Kadire, Gloria Tran, Mahmoud Rahal, Satimai Aniwan, Joseph Meserve, Aaron Weiss, Jenna L. Koliani‐Pace, James P. Campbell, David Faleck, Adam Winters, and Shreya Chablaney
REFERENCES
- 1. Mao EJ, Hazlewood GS, Kaplan GG, et al. Systematic review with meta‐analysis: comparative efficacy of immunosuppressants and biologics for reducing hospitalisation and surgery in Crohn's disease and ulcerative colitis. Aliment Pharmacol Ther. 2017;45:3‐13. [DOI] [PubMed] [Google Scholar]
- 2. Dulai PS, Siegel CA, Colombel JF, et al. Systematic review: monotherapy with antitumour necrosis factor alpha agents versus combination therapy with an immunosuppressive for IBD. Gut. 2014;63:1843‐1853. [DOI] [PubMed] [Google Scholar]
- 3. Colombel J‐F, Sands BE, Rutgeerts P, et al. The safety of vedolizumab for ulcerative colitis and Crohn's disease. Gut. 2017;66:839‐851. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Meserve J, Aniwan S, Koliani‐Pace JL, et al. Retrospective analysis of safety of vedolizumab in patients with inflammatory bowel diseases. Clin Gastroenterol Hepatol. 2018;17:1533‐1540.e2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Singh S, Fumery M, Sandborn WJ, et al. Systematic review and network meta‐analysis: first‐ and second‐line biologic therapies for moderate‐severe Crohn's disease. Aliment Pharmacol Ther. 2018;48:394‐409. [DOI] [PubMed] [Google Scholar]
- 6. Dreyer NA, Bryant A, Velentgas P. The GRACE checklist: a validated assessment tool for high quality observational studies of comparative effectiveness. J Manag Care Spec Pharm. 2016;22:1107‐1113. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Berger ML, Sox H, Willke RJ, et al. Good practices for real‐world data studies of treatment and/or comparative effectiveness: recommendations from the joint ISPOR‐ISPE Special Task Force on real‐world evidence in health care decision making. Value Health. 2017;20:1003‐1008. [DOI] [PubMed] [Google Scholar]
- 8. von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370:1453‐1457. [DOI] [PubMed] [Google Scholar]
- 9. Yao XI, Wang X, Speicher PJ, et al. Reporting and guidelines in propensity score analysis: a systematic review of cancer and cancer surgical studies. J Natl Cancer Inst. 2017;109:djw323. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Dulai PS, Singh S, Jiang X, et al. The real‐world effectiveness and safety of vedolizumab for moderate‐severe Crohn's disease: results from the US VICTORY Consortium. Am J Gastroenterol. 2016;111:1147‐1155. [DOI] [PubMed] [Google Scholar]
- 11. Dulai PS, Thompson KD, Blunt HB, et al. Risks of serious infection or lymphoma with anti‐tumor necrosis factor therapy for pediatric inflammatory bowel disease: a systematic review. Clin Gastroenterol Hepatol. 2014;12:1443‐1451; quiz e88–e89. [DOI] [PubMed] [Google Scholar]
- 12. Austin PC. The performance of different propensity score methods for estimating marginal hazard ratios. Stat Med. 2013;32:2837‐2849. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Austin PC, Schuster T. The performance of different propensity score methods for estimating absolute effects of treatments on survival outcomes: a simulation study. Stat Methods Med Res. 2016;25:2214‐2237. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Austin PC. The use of propensity score methods with survival or time‐to‐event outcomes: reporting measures of effect similar to those used in randomized experiments. Stat Med. 2014;33:1242‐1258. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Wolkewitz M, Cooper BS, Bonten MJM, et al. Interpreting and comparing risks in the presence of competing events. BMJ. 2014;349:g5060. [DOI] [PubMed] [Google Scholar]
- 16. Koller MT, Raatz H, Steyerberg EW, et al. Competing risks and the clinical community: irrelevance or ignorance? Stat Med. 2012;31:1089‐1097. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Pirracchio R, Carone M, Rigon MR, et al. Propensity score estimators for the average treatment effect and the average treatment effect on the treated may yield very different estimates. Stat Methods Med Res. 2016;25:1938‐1954. [DOI] [PubMed] [Google Scholar]
- 18. Austin PC. Optimal caliper widths for propensity‐score matching when estimating differences in means and differences in proportions in observational studies. Pharm Stat. 2011;10:150‐161. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Austin PC, Stuart EA. Optimal full matching for survival outcomes: a method that merits more widespread use. Stat Med. 2015;34:3949‐3967. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. McCaffrey DF, Ridgeway G, Morral AR. Propensity score estimation with boosted regression for evaluating causal effects in observational studies. Psychol Methods. 2004;9:403‐425. [DOI] [PubMed] [Google Scholar]
- 21. Hernan MA, Hernandez‐Diaz S, Werler MM, et al. Causal knowledge as a prerequisite for confounding evaluation: an application to birth defects epidemiology. Am J Epidemiol. 2002;155:176‐184. [DOI] [PubMed] [Google Scholar]
- 22. Brookhart MA, Schneeweiss S, Rothman KJ, et al. Variable selection for propensity score models. Am J Epidemiol. 2006;163:1149‐1156. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Adelson JL, McCoach DB, Rogers HJ, et al. Developing and applying the propensity score to make causal inferences: variable selection and stratification. Front Psychol. 2017;8:1413. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Patrick AR, Schneeweiss S, Brookhart MA, et al. The implications of propensity score variable selection strategies in pharmacoepidemiology: an empirical illustration. Pharmacoepidemiol Drug Saf. 2011;20:551‐559. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Wyss R, Girman CJ, LoCasale RJ, et al. Variable selection for propensity score models when estimating treatment effects on multiple outcomes: a simulation study. Pharmacoepidemiol Drug Saf. 2013;22:77‐85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Green KM, Stuart EA. Examining moderation analyses in propensity score methods: application to depression and substance use. J Consult Clin Psychol. 2014;82:773‐783. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Narula N, Peerani F, Meserve J, et al. Vedolizumab for ulcerative colitis: treatment outcomes from the VICTORY Consortium. Am J Gastroenterol. 2018;113:1345. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Barre A, Colombel JF, Ungaro R. Review article: predictors of response to vedolizumab and ustekinumab in inflammatory bowel disease. Aliment Pharmacol Ther. 2018;47:896‐905. [DOI] [PubMed] [Google Scholar]
- 29. Dulai PS, Osterman MT, Lasch K, et al. Market access analysis of biologics and small‐molecule inhibitors for inflammatory bowel disease among US health insurance policies. Dig Dis Sci. 2019;64:2478‐2488. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Faleck DM, Winters A, Chablaney S, et al. Shorter disease duration is associated with higher rates of response to vedolizumab in patients with Crohn's disease but not ulcerative colitis. Clin Gastroenterol Hepatol. 2019;17:2497‐2505. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Dulai PS, Singh S, Casteele NV, et al. Should we divide Crohn's disease into ileum‐dominant and isolated colonic diseases? Clin Gastroenterol Hepatol. 2019;17:2634‐2643. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Cohen J. Statistical Power Analysis for the Behavioral Sciences. 2nd ed Mahwah, NJ: Lawrence Erlbaum Associates, Publishers; 1988. [Google Scholar]
- 33. Osterman MT, Sandborn WJ, Colombel J‐F, et al. Crohn's disease activity and concomitant immunosuppressants affect the risk of serious and opportunistic infections in patients treated with adalimumab. Am J Gastroenterol. 2016;111:1806‐1815. [DOI] [PubMed] [Google Scholar]
- 34. Singh S, Facciorusso A, Dulai PS, Jairath V, Sandborn WJ. Comparative risk of serious infections with biologic and/or immunosuppressive therapy in patients with inflammatory bowel diseases: a systematic review and meta‐analysis. Clin Gastroenterol Hepatol. 2020;18:69‐81.e3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Lichtenstein GR, Feagan BG, Cohen RD, et al. Serious infection and mortality in patients with Crohn's disease: more than 5 years of follow‐up in the TREAT registry. Am J Gastroenterol. 2012;107:1409‐1422. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Shivaji UN, Sharratt CL, Thomas T, et al. Review article: managing the adverse events caused by anti‐TNF therapy in inflammatory bowel disease. Aliment Pharmacol Ther. 2019;49:664‐680. [DOI] [PubMed] [Google Scholar]
- 37. Koliani‐Pace JL, Singh S, Luo M, et al. Changes in vedolizumab utilization across us academic centers and community practice are associated with improved effectiveness and disease outcomes. Inflamm Bowel Dis. 2019;25:1854‐1861. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Sealed Envelope Ltd . Power calculator for binary outcome non‐inferiority trial [online]. 2012. https://www.sealedenvelope.com/power/binary‐noninferior/. Accessed April 22, 2019
- 39. Panaccione R, Löfberg R, Rutgeerts P, et al. Efficacy and safety of adalimumab by disease duration: analysis of pooled data from Crohn's disease studies. J Crohns Colitis. 2019;13:725‐734. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Ananthakrishnan AN, Sakuraba A, Barnes EL, et al. The benefit of combination therapy depends on disease phenotype and duration in Crohn's disease. Aliment Pharmacol Ther. 2017;46:162‐168. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. O'Neill RT. Secondary endpoints cannot be validly analyzed if the primary endpoint does not demonstrate clear statistical significance. Control Clin Trials. 1997;18:550‐556; discussion 61–67. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1