First-line biologic therapy selection highly impacts ulcerative colitis outcomes: results from the APPETISER study
Article information
Abstract
Background/Aims
We compared the effectiveness of 2 strategies based on the choice of the first biologic owing to either drug efficacy or better acceptability in patients with ulcerative colitis (UC).
Methods
We collected real-world data of all consecutive UC patients ≥ 18 years starting a first line of biologic for active UC with follow-up > 6 months. Patients were considered as receiving either an effectiveness-based strategy (EBS) if they started with infliximab or vedolizumab, or acceptability-based strategy (ABS) if they started with adalimumab or golimumab. Corticosteroid-free clinical remission (PRO2-CFREM) was defined as the absence of bleeding, normalization of stool frequency and no steroid assessed as a binary criterion each month. All comparisons were adjusted using propensity scores.
Results
Overall, 130 patients and 3,355 months were analyzed. The percentage of months spent in PRO2-CFREM within the first 24 months (primary endpoint) was greater in EBS arm than in ABS arm (74.2% vs. 46.6%; P< 0.001), was higher within the first 6 months (60.4% vs. 18.9%), and was not rescued later between months 7 and 12 (73.1% vs. 38.6%), months 13 and 18 (79.9% vs. 57.0%) or M19 and M24 (83.3% vs. 64.6%) (P< 0.001 for all comparisons). EBS group had a lower risk to first-line biologic discontinuation (adjusted hazard ratio [aHR], 6.5; 95% confidence interval [CI], 2.8–15.3: P< 0.001) and a trend for lower risk of colectomy (aHR, 4.5; 95% CI, 0.9–22.3; P= 0.068).
Conclusions
The choice of first-line biologic is highly impacting UC outcome and cannot be fully compensated by later treatments, advocating for the use of the most effective therapeutic option as first-line biologic.
INTRODUCTION
Ulcerative colitis (UC), is a progressive, chronic and disabling disorder that can lead to colectomy, colorectal cancer, disability and altered quality of life [1,2]. In the last decade, the therapeutic armamentarium has been highly enlarged in UC owing to the efficacy of new drugs in this indication. Thus, after failing to conventional therapy such as 5-ASA (5-aminosalicylic acid) or steroids, the choice of the first advanced therapy is now more complex. Except for cost considerations, decision-making should be personalized and should rely on the triptych efficacy-safety-acceptability with individualized weighting according to each patient. The hierarchization of inflammatory bowel disease (IBD) drugs based on efficacy is not clear due to the lack of head-to-head randomized controlled trials. The VARSITY trial was the only study and demonstrated the superiority of vedolizumab over adalimumab in bio-naive patients with moderate-to-severe UC [3]. In this context, indirect comparisons retrieved from network meta-analyses or real-world evidence can be helpful. Network meta-analyses [4,5] suggest higher efficacy of infliximab and vedolizumab compared to full subcutaneous anti-tumor necrosis factor (TNF), i.e., adalimumab and golimumab. However, despite the availability of subcutaneous formulations, intravenous induction and maintenance can be felt as burdensome by the patients as demonstrated by the ACCEPT2 study [6]. Thus, the impact of therapeutic sequencing is a burning question and still unknown in patients with UC. Whether starting with a less effective but safer and more convenient drug is a loss of chance for UC patients or can be rescued by later treatments is key point. We can hypothesize that drug sequencing could be more significant than the choice of the first medication by itself. Then, starting with one of these therapeutic classes with rescue by another therapy among patients experiencing therapeutic failure could lead to similar mid-term and long-term outcomes regardless of the choice of first-line biologic. Confirming this hypothesis could lead to select the treatments according to the safety profile, the patients’ preference and the lower costs for the healthcare system. For attempting to address this question, we designed the APPETISER study aiming at comparing the effectiveness of 2 strategies based on the choice of the first biologic owing to either the drug efficacy (vedolizumab or infliximab based on data from network meta-analyses [4,5] and randomized controlled trials [3]) or better acceptability [6] (adalimumab and golimumab) to induce and maintain steroid-free clinical remission during the first 24 months in patients with UC.
METHODS
1. Ethical Considerations
The study was approved by the Local Ethics Committee (IRB 00013412, CHU de Clermont-Ferrand IRB #1, IRB number 2023-CF074) with compliance to the French policy of individual data protection. All authors had access to the study data and reviewed and approved the final manuscript.
2. Selected Population and Therapeutic Algorithm
From the database of an IBD referral center, we retrieved real-world but well standardized data from medical records of all bio-naive patients with UC >18 years, starting a biological therapy for symptomatic UC (partial Mayo score >2) with a follow-up longer than 6 months. Patients with prior colectomy, unclassified colitis or acute severe colitis were excluded. Patients were included in 2 groups: either effectiveness-based sequencing (EBS) if they started with infliximab or vedolizumab, or acceptability-based sequencing (ABS) if they received adalimumab or golimumab as first-line biologic. Thus, effectiveness-based or acceptability-based terms are referred only for the choice of the first-line therapy. The time to intensify and to start a subsequent new line of treatment as well as the choice of second-line biologic was let at the physician’s discretion but was standardized based on current therapeutic algorithm.
3. Data Collection
The following data were collected and anonymized: patients characteristics (age, smoking, sex), UC characteristics (duration, extent), prior UC medications, therapeutic sequencing, (dose, interval, concomitant immunosuppressants), clinical (partial Mayo score, patients-reported outcome [PRO2]), biological (C-reactive protein, fecal calprotectin), and endoscopic activity (endoscopic Mayo score). In addition, histological activity (Nancy index) was collected if available.
4. Endpoints
The primary endpoint was symptomatic remission defined according to PRO2 as the absence of bleeding, normalization of transit (Mayo stool frequency subscore=0) and no steroid. PRO2-symptomatic remission was defined as a binary criterion each month (the statistical unit being the month and not the patient). The secondary endpoints concerned the analysis by patient and were in corticosteroid-free clinical remission (CFREM; defined as partial Mayo score ≤ 2 without steroids) at week 12, endoscopic remission (CFREM + endoscopic Mayo score ≤ 1) at week 12, time to first biologic discontinuation, time to first hospitalization and time to colectomy.
5. Sample Size Calculation
We aimed to highlight an absolute difference of 15% (PRO2-symptomatic remission) for a 2-sided type I error at 0.05 and a statistical power at 90%. This difference (58% vs. 73%) is based on the data from the ACCEPT 2 study [6] and clinical relevance. To take into account 2 potential issues (1) missing data (expected less than 5%) and (2) the intra-class correlation coefficient (ICC) measuring the between and within patient variability (several measures for a same patient as the statistic unit is month; ICC fixed at 0.10), we assessed that 700 and 350 months will be necessary for the EBS and ABS groups, respectively. Expecting a median follow-up of at least 12 months (12 periods of 4 weeks), a total number of at least 90 patients (60 in EBS group and 30 in ABS group) were required.
6. Statistical Analyses
Patients’ characteristics were expressed as mean ±standard deviation or median and interquartile range for continuous data. The assumption of normality was assessed by using the Shapiro-Wilk test. As indication bias is frequent in retrospective studies (the choice of treatment is generally influenced by patients’ characteristics), we used a propensity score to balance for these differences and recreate, as much as possible, the gold standard of a randomized trial. Inverse probability of treatment weighting was used by assigning each participant an inverse weighting of the probability of being included in EBS or ABS groups, estimated by the propensity score in order to make the 2 groups comparable. The validity of the matching was tested by analyzing the standardized differences. Considering the characteristics of the subjects at baseline and clinical relevance, the propensity score model included the following variables: sex, UC extent (E3 vs. E2/E1), concomitant immunosuppressant or steroids and disease severity (partial Mayo score ≥ 6 or not at baseline). The validity of the weighting was tested by analyzing the standardized differences (|d|), with |d| >0.2 considered to be an imbalance. Then, the comparisons between groups for continuous variables were performed using Student t test or Mann-Whitney test when the assumptions of t test were not met. The homoscedasticity was studied using Fisher-Snedecor test. For categorical data, chi-square and Fisher exact tests were applied. Time-to-event censored data were estimated using Kaplan-Meier method. Log-rank test was used to determine prognostics factors associated to the censored outcomes. The proportional-hazard hypothesis was studied using Schoenfeld’s test and plotting residuals. Cox proportional hazards model for time-to-event analysis to calculate adjusted hazard ratios (HRs). The longitudinal analysis of PRO2-symptomatic remission was conducted using last observation carried forward method, clinical relevance, and random-effects model, more precisely logistic mixed model. Time, group and their interaction were considered as fixed effects whereas patient was treated as random-effect in order to model between and within subject variability. Statistical analyses were performed using Stata 15 software (StataCorp, College Station, TX, USA). All tests were 2-sided, with a type I error set at 0.05.
RESULTS
1. Baseline Characteristics of Patients
Overall, 130 patients were included including 60 patients in the EBS group (50 infliximab and 10 vedolizumab) and 70 in the ABS group (48 adalimumab and 22 golimumab) (Table 1). Briefly, the populations had similar characteristics at baseline apart from concomitant immunosuppressant more frequent in EBS group (56.6% vs. 22.9%, P=0.01). Regarding first line of biological therapy in EBS group, infliximab was intensified in 94.0% of the patients (47/50) while only 10.0% (1/10) received intensified dose of vedolizumab (300 mg every 4 weeks). Among the patients included in ABS group, adalimumab as first-line biological therapy was dose escalated in 81.3% (39/48) of the patients to 40 mg every week or 80 mg every other week including 10.4% (5/48) to 80 mg every week. First line of biological therapy with golimumab was intensified to 100 mg every 4 weeks in 63.6% (14/22) of the patients with 22.7% (5/22) subsequently dose escalated to 100 mg every other week.
Baseline Characteristics of Patients with UC Included in APPETISER Study
Overall, second-line treatment was started in 26.7% (16/60) of patients in EBS group due to failure of the first-line therapy (adalimumab in 7/16, 43.8%; vedolizumab in 3/16, 18.8%; ustekinumab in 3/16, 18.8%) and 87.1% of patients in ABS group due to failure of the first-line therapy (infliximab in 33/62, 53.2%; vedolizumab 19/62, 30.6%). It means that a vast majority of patients from the ABS group received either infliximab or vedolizumab as second-line therapy (52/62, 83.8%). The details of therapeutic sequences are given in Fig. 1.
(A) Flowchart illustrating the therapeutic sequences in the APPETISER study among efficacy-based strategy. (B) Flowchart illustrating the therapeutic sequences in the APPETISER study among acceptability-based strategy. aPrimary failure (n=4), Secondary loss of response (n=6), Intolerance (n=4); bPrimary failure (n=23), Secondary loss of response (n=10), Intolerance (n=4), other (n=4); cPrimary failure (n=14), Secondary loss of response (n=5), Intolerance (n=1). IFX, infliximab; VDZ, vedolizumab; ADA, adalimumab; GOL, golimumab; UST, ustekinumab; TOF, tofacitinib; FIL, filgotinib.
2. Comparative Effectiveness after Induction Therapy between the 2 Groups (Per Patient Analyses)
After adjustment, the rate of CFREM at week 12 was significantly higher in the EBS group (70.8%) than in the ABS group (33.0%) (P<0.001) (Fig. 2). The rate of endoscopic remission at week 12 was 22.1% and 10.0% in EBS and ABS arms, respectively (P=0.086).
Corticosteroid-free clinical remission after induction therapy (week 12) in patients included in efficacy-based or acceptability-based strategies.
3. Time Spent in PRO2-Symptomatic Remission within the first 24 Months (Per Month Analyses)
For the primary endpoint, 3,355 months were analyzed (1,180 in EBS arm and 1,175 in ABS arm). After propensity score analyses, the percentage of months spent in PRO2-symptomatic remission, within the first 24 months (primary endpoint) was greater in EBS arm than in ABS arm (74.2% vs. 46.6%; P<0.001) (Fig. 3A). The percentage of months spent in PRO2-symptomatic remission was higher in the EBS arm within the first 6 months (60.4% vs. 18.9%), and was not rescued later: between months 7 and 12 (73.1% vs. 38.6%), months 13 and 18 (79.9% vs. 57.0%) or months 19 and 24 (83.3% vs. 64.6%) (P<0.001 for all comparisons) (Fig. 3B). The detail of time spent in PRO2-symptomatic remission according to each first-line therapy is given in Supplementary Table 1.
(A) Percentage of months spent in PRO2-symptomatic remission within the first 24 months in patients included in efficacy-based or acceptability-based strategies. (B) Percentage of months spent in PRO2-symptomatic remission within the first 24 months by semester in patients included in efficacy-based or acceptability-based strategies.
4. Long-term Effectiveness (Per Patient Analyses)
The median follow-up was 70.6 months (interquartile range, 43.1–101.6 months). During the follow-up of the study, 79 patients (60.8%) had to discontinue their first biological therapy due to therapeutic failure including 17 in the EBS group (28.3%) and 62 among ABS group (88.6%). After propensity score analyses, EBS group had a lower risk of first-line biologic discontinuation (adjusted HR [aHR], 6.5; 95% confidence interval [CI], 2.8–15.3; P<0.001) (Fig. 4A). In addition, 14 patients (11.2%) had to be hospitalized owing to symptomatic UC including 4 (6.4%) and 10 (14.9%) among EBS and ABS arms, respectively. After adjustment, we observed a trend between the 2 groups regarding the time to the first hospitalization (aHR, 4.4; 95% CI, 0.9–22.9; P=0.06) (Fig. 4B). During the follow-up, the rate of colectomy was limited (12 patients, 9.8%). Colectomy was required in 7 (12.3%) and 5 patients (7.6%) in EBS and ABS patients. After propensity score analysis, we observed a trend for lower risk of colectomy (aHR, 4.5; 95% CI, 0.9–22.3; P=0.068) compared to ABS group (Fig. 4C).
(A) Kaplan-Meier curve illustrating first biologic discontinuation-free survival in patients with ulcerative colitis included in efficacybased or acceptability-based strategies. (B) Kaplan-Meier curve illustrating hospitalization-free survival in patients with ulcerative colitis included in efficacy-based or acceptability-based strategies. (C) Kaplan-Meier curve illustrating colectomy-free survival in patients with ulcerative colitis included in efficacy-based or acceptability-based strategies.
DISCUSSION
In the real-world evidence APPETISER study, we confirmed that therapeutic sequences starting with infliximab and vedolizumab are more effective than adalimumab or golimumab to induce remission, and are associated with better long-term outcomes, highlighting that the choice of first-line biological therapies is highly impacting the outcome of UC. Our data suggest that sequencing starting with less effective but more acceptable treatment is a loss of chance for the patients with UC as it cannot be rescued by later treatments.
To investigate the impact of therapeutic sequencing in UC, we took advantage from real-world data applying usual statistical methods including propensity score analyses. Of note, we paid particular attention to the choice of items included in the propensity score as it is a critical point for reliability of the results [7-9]. We used a pragmatic endpoint expressed as the percentage of months spent in PRO2-symptomatic remission to approach the concept of continuous remission in the same way as previous [10] or ongoing studies (STARTER project from the GETAID). The definition of our 2 groups is based on data from randomized controlled trials such as VARSITY3 but also network meta-analyses [4,5] and real-world [11,12] data showing that infliximab and vedolizumab are more effective than adalimumab and golimumab in biologics-naive patients with UC. While data from network meta-analyses and pooled data from 3 randomized controlled trials reported close efficacy between infliximab and vedolizumab in biologic-naive patients with UC [13], real-world evidence could suggest higher effectiveness of vedolizumab compared to infliximab therapy [14].
In our study, we observed very high rate of CFREM at week 12 in patients included in EBS strategy as almost 3 quarters of the patients achieved this endpoint. This result is mostly due to infliximab effectiveness as 83.3% (50/60) of the patients were treated with infliximab as first-line biologic among EBS group. Infliximab was the first-in-class anti-TNF therapy and dramatically change the management of patients with UC owing to its efficacy to induce remission as shown by the landmark studies ACT and SUCCESS [15,16]. However, the observed rate of remission seems higher in our study. It could be partly due to the real-world setting, which usually leads to better results of effectiveness owing to less selected patients and dose or interval optimization. In another hand, results from different network meta-analyses and real-world cohorts indicate that adalimumab and golimumab are less effective than infliximab as first-line biologics [4,5,12]. We observed the same trend regarding the achievement of clinical and endoscopic remission with more than 2-fold higher effectiveness in patients treated with infliximab or vedolizumab compared to those receiving adalimumab and golimumab, which is in line with results from randomized controlled trials and network meta-analyses [3,15-19].
Regarding our result on continuous remission reflected by the time spent in PRO2-symptomatic remission, the superiority of EBS strategy within the first 6 months was expected due to the high difference in effectiveness after induction phase. However, one important result is that the time spent in remission was still higher throughout the first 24 months despite the use of second-line therapies in ABS arm. Even though the difference reduced over time, it seems that we cannot recapture the same level of time spent in remission. It is of particular meaning for practice suggesting that the choice of first-line biologic determine the outcome of UC despite the use of rescue therapies such as infliximab or vedolizumab ( >80% in our study). Although all biological therapies are less effective in biologics-exposed patients with UC, vedolizumab seems to be more impacted than the others [5,9]. However, Hupé and colleagues suggested that vedolizumab could be a better option than infliximab after failing to a first anti-TNF agent [20]. Despite this, infliximab remains an option as second-line anti-TNF therapy [21] but the type of failure should be taken into account [22]. Two main hypotheses can be raised to explain lower efficacy of second-line medications [5]: either therapeutic failure is predominantly a marker of severity regardless of the sequence or the use of the first-line biologic is dramatically changing the immune system response [23] limiting the efficacy of a second therapy meaning that therapeutic sequencing is crucial and highly condition the long-term outcomes of patients with UC. In the same way, pejorative endpoint such as treatment failure, or negative long-term outcomes such as hospitalization or even colectomy will be prompter to occur among patients starting with less effective first-line advance therapy. It is in line with data from a nationwide Danish cohort study reporting a higher risk of UC-related hospitalization in patients treated with adalimumab compared to infliximab in biologic-naive patients with UC [24].
The main limitations of the APPETISER study are the retrospective and monocenter design and the low use of other biological therapy or small molecules such as ustekinumab and JAK (Janus kinase) inhibitors during the study period even though our goal was to conceptually investigate whether therapeutic sequencing has a significant impact on UC outcome, and not a specific sequence of treatments. The heterogeneity of dose optimization could also have slightly impacted some secondary endpoints such as time to first biologic discontinuation. In addition, even though taking into account potential confounders is more important than the sample size by itself as aforementioned, our sample size may have limited the power to detect differences in patient-centered outcomes, particularly endoscopic remission.” Several strengths have to be underlined such as the inclusion of consecutive patients with daily practice standardized evaluation and high-quality collected data anticipating the future use for real-world evidence research and methodological expertise in comparing real-world data [7,9,25]. Besides, our study provides clear and meaningful results that can be translated into daily practice.
In conclusion, our data highlight that the choice of first-line biological therapies is highly impacting the outcome of UC and cannot be fully compensated by later treatments. The results of the APPETISER study advocate for the use of the most effective therapeutic option as first line of biologic. However, these data should be confirmed in dedicated randomized controlled trial such as the STARTER project.
Notes
Funding Source
The authors received no financial support for the research, authorship, and/or publication of this article.
Conflict of Interest
Buisson A has received consulting fees from AbbVie, Amgen, Arena, Biogen, Celltrion Healthcare, CTMA, Galapagos/Alfa-Sigma, GutyCare/Resilience Janssen, Lilly, MSD, Nexbiome, Pfizer, Roche, Sandoz, Takeda and Tillotts. Lecture fees from AbbVie, Amgen, Biogen, Celltrion Healthcare Galapagos/Alfa-Sigma, Janssen, Lilly, Mayoli-Spindler, MSD, Nordic Pharma, Norgine, Pfizer, Roche, Takeda, Tillotts and Vifor Pharma. Research grant from AbbVie, Celltrion Healthcare, Janssen, Lessaffre, Lilly, Pfizer, Sandoz and Takeda. All other authors declare no conflicts of interest.
Data Availability Statement
Data are available from the corresponding author upon reasonable request.
Author Contributions
Conceptualization: M’Baye D, Hupé M, Pereira B, Buisson A. Data curation: M’Baye D, Dodel M, Bazoge M. Formal analysis: M’Baye D, Bazoge M, Pereira B, Buisson A. Investigation: Dodel M, Bazoge M. Methodology: Hupé M, Pereira B, Buisson A. Project administration: Dodel M. Supervision: Pereira B, Buisson A. Validation: Hupé M, Dodel M, Bazoge M, Pereira B. Writing–original draft: M’Baye D, Buisson A. Writing–review & editing: Hupé M, Dodel M, Bazoge M, Pereira B. Approval of final manuscript: all authors.
Additional Contributions
We thank CHU Clermont-Ferrand (DRCI) for its recurrent support.
Supplementary Material
Supplementary materials are available at the Intestinal Research website (https://www.irjournal.org).
Supplementary Table 1.
Percentage of Months Spent in PRO2-Symptomatic Remission within the First 24 Months and by Semester in Patients Included in Efficacy-Based or Acceptability-Based Strategies