The Effect of Early vs Delayed Initiation of Adalimumab on Remission Rates in Patients With Crohn’s Disease With Poor Prognostic Factors: The MODIFY Study

Abstract Background Data on the effectiveness of anti-tumor necrosis factor medications in patients with Crohn’s disease (CD) with poor prognostic factors (PPFs) are scarce. This study aimed to generate real-world evidence on the effect of early (≤24 months after diagnosis) vs delayed (>24 months) initiation of adalimumab (ADL) on the 26-week remission rate (Harvey–Bradshaw Index ≤4) in these patients. Methods This multicentre, retrospective, chart review study performed in 10 Greek hospitals enrolled adult patients with moderate to severe CD (Harvey–Bradshaw Index ≥8) with ≥3 PPFs who were initiated on ADL ≥12 months before enrollment. A sample size of 164 patients (early:delayed cohort allocation ratio, 30:70) was required to address the primary endpoint. Results Eligible patients (n = 171) were consecutively enrolled. In the early vs delayed cohorts, the 26-week remission rates (off-steroids) using the last-observation-carried-forward imputation method were 60.7% (37/61) vs 47.2% (50/106), respectively (Δ = 13.5%, P = .044). The respective remission rates were 61.2% vs 42.4% among anti-tumor necrosis factor-naive patients (P = .023) and 58.3% vs 53.2% among anti-tumor necrosis factor-experienced patients (P = .374). The 52-week remission rates using as-observed data were 78.8% and 60.3%, and the intestinal resection rates were 6.5% and 11.9% in the early vs delayed ADL cohorts, respectively. Conclusions Patients with CD with PPFs who received early vs delayed treatment with ADL achieved higher clinical response and remission rates. This effect was more pronounced in those patients who were bio-naive and steroid-dependent/refractory with concurrent extraintestinal manifestations than those who were not.


Introduction
Crohn's disease (CD) is a chronic disease with dynamic progression towards complicated phenotypes, 1 resulting in disabling symptoms. 2 Identification of prognostic factors associated with certain long-term poor outcomes has been the focus of several epidemiologic studies. [3][4][5][6][7] The identification and validation of prognostic factors is a major objective in clinical research, not only as a means to predict the natural history of CD, but also to apply highly individualized therapeutic strategies that would yield optimized disease outcomes by avoiding both over-and under-treatment. 8,9 However, to halt the progressive bowel damage in CD beyond the identification of poor prognostic factors (PPFs) early treatment with disease-modifying drugs, such as antitumor necrosis factor (TNF) agents, is also required. 10,11 Early CD is defined as an active, inflammatory disease with objective signs of disease activity but no bowel damage and a disease duration of <2 years. 12 This definition has been based on the paradigm of rheumatoid arthritis, 13,14 taking into consideration the fact that anti-TNFs [adalimumab (ADL), infliximab, certolizumab pegol] are more effective in patients with disease duration of <2 years.
Even though it is well recognized that identifying PPFs would allow individualized, early anti-TNF intervention, data on the effectiveness of anti-TNFs in patients with CD with poor prognosis are scarce. This study aimed to generate real-world evidence on the effect of early vs delayed initiation of ADL in patients with PPFs. Further, the potential association between early vs delayed initiation of ADL in the CD course and prevention of poor long-term outcomes was examined.

Study Design and Setting
This single-country, multicentre, retrospective, chart review study was performed by gastroenterologists specialized in inflammatory bowel disease (IBD) in 10 IBD centers of public hospitals in Greece. Sites were selected through a documented and structured feasibility process that accounted for physicians' qualifications, previous participation and experience in similar studies, number of potentially eligible patients, data availability, and staff resourcing. To represent variations in current real-world patterns of care, research sites were recruited from various geographic regions in Greece, taking into consideration the regional setting and type of healthcare site/institution (~31% of the patients recruited from academic and the remaining 69% from nonacademic sites) as well as the hospital activity.
The study involved secondary data collection from medical charts of patients with CD treated with ADL by means of an electronic web-based data capture system. Medical charts were reviewed and assessed through a process of consecutive sampling that followed reverse chronologic order based on the date of ADL treatment onset. Beginning with the most recently ADL-initiated patients (ie, 12 months before selection process onset) and then moving retrospectively, patients were assessed for study-specific eligibility criteria. All eligible patients who were alive at the time of chart abstraction initiation were contacted by the site staff to provide consent on the use of their medical charts. If the eligible patient could not be reached during the initial telephone contact, site staff attempted telephone contact 3 more times before considering the eligible patient to be not interested in study participation. Eligible patients who attended the participating clinical sites within the context of a routine clinical visit during the selection process period were also considered for inclusion in the study.
The study look-back period for data collection comprised (i) the pre-ADL initiation period, spanning from the date of CD diagnosis to the day before ADL treatment initiation and (ii) the period after ADL initiation, which began on the date of ADL treatment initiation and ended at loss to follow-up, chart abstraction initiation (for patients for whom ADL was ongoing at enrollment), a switch to another treatment, or at 3 months after ADL discontinuation, whichever came first.

Study Population
The study population involved adult male and female consenting patients with moderately to severely active CD [Harvey-Bradshaw Index (HBI) score ≥8], 15 who were naive or experienced to biologic therapy at initiation of ADL and who had ≥3 PPFs at CD diagnosis. The rationale for requiring the presence of ≥3 PPFs was based on the European Crohn's and Colitis Organisation guidelines advising early treatment with thiopurines and/or biologics when two or more predictors are present. 16 PPFs included ileal or ileocolonic location, age ≤40 years, active smoking, extensive and deep ulceration, severe (as per physician judgment) rectal and/or perianal disease, and upper gastrointestinal involvement. Patients had initiated treatment with ADL, as per the approved label, ≥12 months before enrollment in the study and after it was launched in the Greek market (June 2007). Patients should also have had sufficient available medical records for data abstraction to meet the objectives of the study. Patients with a history of CD-related intestinal resection before initiation of ADL, those who had participated in any clinical trial with any investigational product/ intervention during treatment with ADL, as well as female patients who had become pregnant during ADL treatment were excluded.
Patients were stratified into 2 cohorts, according to the length of CD duration at ADL initiation. Those with CD duration of ≤24 months before initiation of ADL were stratified in the early cohort; those with disease duration of >24 months were placed in the delayed cohort. The study population was recruited through the implementation of a competitive recruitment strategy, until the planned sample size was reached, at a targeted allocation ratio of 30:70 between the early and delayed cohorts.

Study Outcomes
The primary objective of the study was to examine the effect of early vs delayed ADL initiation on the 26 ± 4-week postbaseline clinical remission rate by estimating the difference in the remission (HBI score ≤4) rates (off-steroids) between the 2 cohorts. Baseline was defined as the date of ADL treatment initiation (first injection) or the closest time point before that date. Secondary study outcomes included (i) the difference between the early and delayed cohorts in the 26 ± 4-week offsteroids clinical response rate, defined as a ≥3-point decrease in the HBI score; (ii) the relative risk of intestinal resection (excluding surgeries for perianal disease) among patients in the early and delayed cohorts; (iii) the relative risk of CD progression from an inflammatory to a stricturing and/or penetrating phenotype among patients in the 2 cohorts; and (iv) the difference between the 2 cohorts in the proportion of patients remaining on treatment at week 52 ± 4 (drug survival rate). The following are PPFs of interest and their definitions: active smoking, defined as smoking >7 cigarettes per week for ≥6 months immediately before CD diagnosis; extensive and deep ulceration, defined as large coalescent and deep ulcerations covering more than 10% of the mucosal area of ≥1 segment of the rectocolon; and upper gastrointestinal involvement, defined as esophageal, gastric, duodenal or jejunal inflammation due to CD.
The following are the baseline factors of interest: (a) the presence of extensive intestinal mucosal inflammation, defined as intestinal CD affecting >100 cm regardless of location of the inflammation; (b) being steroid-dependent, defined as either being unable to reduce steroids below the equivalent of prednisolone 10 mg/day (or budesonide below 3 mg/ day) within 3 months of starting steroids without recurrent active disease or having had a relapse within 3 months of stopping steroids; and (c) being steroid-refractory, defined as having active disease despite treatment with prednisolone up to 1 mg/kg/day for a period of 4 weeks.

Sample Size Estimation
The sample size calculation was based on the study's primary endpoint. To detect a difference of 20.0% between the early vs delayed treatment cohorts with 80% power and an assumed 40.0% rate in the early treatment group, 17, 18 a sample size of 164 was required, assuming a study size allocation ratio of 30:70 (early:delayed) and a 10% non-evaluable rate, using a 1-sided chi-square test at a 0.05 significance level. The sample size was calculated with SAS statistical analysis software version 9.4 using the PROC POWER TwoSampleFreq statement with the TEST = PChi and NFRACTIONAL options (SAS Institute, Cary, NC, USA).

Statistical Analysis
Statistical analysis was performed using SAS software version 9.4. The normality of distribution was examined using the Shapiro-Wilk test. Data are presented as mean (SD) when normally distributed and as median [interquartile range (IQR)] when not. Regarding binomial proportions, 95% Wald CIs have been estimated. The difference between the 2 cohorts in the HBI remission and clinical response rates were estimated, and the statistical significance of the difference was examined with a Wald-based z-test statistic (1-sided test; a = 5%). Analyses of 26-week HBI remission rates and clinical response rates have been performed using the last-observation-carried-forward (LOCF), nonresponder imputation (NRI) methods, and as-observed data. To evaluate potential differences in the primary endpoint among selected subgroups, the difference in the HBI ≤4 attainment rate (LOCF imputation method) between the 2 study cohorts was examined and is presented separately in the subgroups per prior anti-TNF exposure (anti-TNF-naive, anti-TNFexperienced), steroid dependence/refractoriness, and the presence of extraintestinal manifestations at baseline. Differences between the values of continuous variables in 2 time periods were evaluated using a parametric statistical test (ie, paired t-test) for normally distributed data or its nonparametric analog (ie, Wilcoxon signed-rank test). Kaplan-Meier analysis was performed for the estimation of the median time to intestinal resection and progression to a stricturing/penetrating phenotype. Differences in the survival distributions between the 2 cohorts were examined with the log-rank test. The effect of study cohort (early vs delayed) on the rates of intestinal resection/progression to a stricturing/penetrating phenotype was evaluated using hazard ratios (HRs) and 95% CIs estimated from Cox proportional hazards models.
Finally, longitudinal analysis of C-reactive protein (CRP) over time was performed to assess the effect of early vs delayed ADL onset on CRP measurements obtained between ADL treatment onset and the end of the observation period. Repeated measures of CRP were used as the dependent variable and study cohort (early, delayed) time elapsed from ADL treatment onset to assessment (in weeks) and their interaction as independent variables. The Akaike information criterion (AIC) was used to compare linear mixed-effects/marginal models with different covariance structures; the model with the smallest AIC value was chosen. The interaction between cohort and time was not statistically significant, so inference was based on a model (fitted using the same covariance structure) including only the main effects. Unless specifically stated, statistical tests were 2-sided and were performed at a 0.05 significance level.

Ethical Considerations
The study was designed and conducted in compliance with all applicable local laws and regulations, the Good Pharmacoepidemiology Practices of the International Society for Pharmacoepidemiology and the ethical principles laid down in the Declaration of Helsinki. The study was approved by the institutional review boards of the participating hospital sites. All patients provided written informed consent.
The overall median (IQR) study look-back period between CD diagnosis and the end of the study observation period was 45.5 (28.8-75.5) and 148.2 (96.0-203.5) months for the early and delayed cohorts, respectively. The median (IQR) observation period after ADL treatment initiation was 37.7 (19.9-61.9) and 38.5 (22.4-70.9) months, respectively.

Poor Prognostic Factors at CD Diagnosis
A median (IQR) of 3 (3-4) of the protocol-specified PPFs were present in each cohort at diagnosis (Figure 2). In particular, 58.1% of the patients in the early cohort had 3, 33.9% had 4, and 8.1% had 5 PPFs. In the delayed cohort, 56.0% had 3, 35.8% had 4, and 8.3% had 5 or 6 PPFs. The 3 most common PPF in both cohorts were ileal or ileocolonic disease location, age ≤40 years, and active smoking ( Figure 2).

Baseline Characteristics
Characteristics of the early and delayed cohorts at baseline (ADL initiation) are presented in Table 1. At baseline, a median (IQR) of 5.3 (3.0-13.8) and 95.8 (48.0-140.7) months had elapsed from CD diagnosis to ADL initiation for the early and delayed cohorts, respectively. The 2 cohorts were not statistically different regarding essential disease characteristics (chi-square test, P > .05; Figure 3). Extraintestinal manifestations were present in 37.1% of the patients in the early and 38.0% of the patients with available data in the delayed ADL cohort. The most common extraintestinal manifestations (ie, in at least 3 patients) in the early ADL cohort were axial At least one clinically significant (according to physician's judgment) past or ongoing disease or surgery (other than those related to CD) was reported for 19.4% and 18.3% of the patients in the early and delayed cohorts, respectively. Before ADL initiation, 2 patients (3.2%) in the early cohort and 3 (2.8%) in the delayed cohort had undergone CD-related surgeries (seton placement, duodenal ulcer repair, and abdominal operation) other than intestinal resection.

CD Treatment Characteristics
In the early and delayed cohorts, 79.0% and 100% of patients, respectively, had received ≥1 therapeutic regimen before ADL initiation ( Table 2). In total, 19.4% of the patients in the early ADL cohort and 45.9% of those in the delayed ADL cohort had received prior biologic therapy, which comprised of anti-TNF agents in all cases. The main reasons for discontinuation of prior anti-TNF therapy in the early ADL cohort were adverse event/intolerance (in 66.7% of the cases) and
Patients in the early group who were bio-naive had extraintestinal manifestations and steroid-dependent/refractory CD at baseline achieved higher 26-week off-steroids remission rates compared with their counterparts in the delayed group (LOCF imputation; Figure 4B-D).
Using the more strict remission criterion of HBI <4, the 26week off-steroids remission rates were 55.7% (34/

HBI Scores and CRP Values Throughout the Study
In both the early and the delayed ADL cohorts, the baseline HBI score significantly decreased at all post-baseline timepoints (4-, 12-, 26-, and 52-week post-baseline), con-  Concerning CRP values, a marginal model with the main effects of time elapsed from ADL onset, ADL cohort, and their interaction, did not reveal a significant interaction betweenstudy cohort and time (P = .416). As a result, a marginal model including only the main effects was fitted according to which time elapsed from ADL initiation was statistically significant [specifically, CRP was estimated to decrease by 0.004 (95% CI, −0.008, −0.000; P = .026) units per week after the start of ADL treatment], but early vs delayed initiation of ADL treatment was not.

CD-Related Intestinal Resection and Progression From an Inflammatory to a Stricturing and/or Penetrating Phenotype
During the study look-back observation period, the CDrelated intestinal resection rates (for reasons other than perianal complications) were 6.5% (4/62) and 11.9% (13/109) in the early and delayed cohorts, respectively, and the rates of CD progression to a complicated (stricturing and/or penetrating) phenotype were 7.7% (3/39) and 11.4% (8/70). The median Kaplan-Meier estimated time to first intestinal resection was not reached in either cohort (see Supplementary Figure 1A for the Kaplan-Meier curve). The HR of the delayed vs early cohort was 1.81 (95% CI, 0.59-5.56; P = .299). The median Kaplan-Meier estimated time to progression to a complicated phenotype was 105.2 months (95% CI, 80.1-105.2) in the early cohort but was not estimable in the delayed cohort (see Supplementary Figure 1B for the Kaplan-Meier curve). The HR of the delayed vs early cohort was 1.33 (95% CI, 0.35-5.09; P = .677).

Discussion
This study has demonstrated that initiation of ADL within the first 2 years following diagnosis in adult patients with moderate to severe CD with at least 3 PPFs resulted in a significantly higher remission rate compared with a delayed ADL introduction.
The effect of disease duration on the efficacy/effectiveness of biologic treatments in patients with CD has been addressed by several studies. [18][19][20][21][22] However, to the best of our knowledge, this is the first evidence derived from a CD study in a uniform population of White patients with poor prognosis, with another study examining the effect of early anti-TNF introduction in an exclusively Asian population with PPFs. 23 In addition to the identified demographic, behavioral, diseasespecific, comorbid, and genetic prognostic factors, this study involved those indicated by the Annual Exchange on the Advances in IBD (IBD Ahead) 2014 educational program, which included a comprehensive literature review to identify prognostic factors related to long-term outcomes and develop evidence-weighted summary statements. 19 Smoking, 6,7,24 young age at diagnosis, 2,7,25,26 ileal disease location, [26][27][28][29] deep and extensive ulceration, 30 severe rectal and/or perianal disease, 2 and upper gastrointestinal involvement, 31 were recognized as predictive for both increased surgery needs and change of CD behavior from inflammatory to stricturing and/ or penetrating phenotypes.
Analysis of the study's primary endpoint of the 26-week remission rates (off-steroids) revealed a statistically significant difference between the early and delayed cohorts. This was more pronounced when using a more stringent definition of remission as the target, indicating the potential of early treatment to achieve even more stringent therapeutic goals. These findings are in line with a subgroup analysis of the CHARM trial, indicating that initiation of ADL within <2 years vs ≥2 years of disease onset offers a significant benefit in remission rates at 26 weeks. 18 Similarly, in the certolizumab PRECISE 2 trial, the 26-week remission rates were inversely related to CD duration. 32 Of note, the populations recruited in these randomized clinical trials do not represent a homogenous sampling of patients with poor prognosis, whereas our study does. Similarly to week 6, higher remission rates in the early vs delayed cohort observed were observed at week 52. Such a difference between early and delayed initiation of ADL is also evident in the 56-week remission rates of the CHARM/ADHERE trials of patients with CD regardless of the presence of PPFs at baseline. 18 Not unexpectedly, in our study, the difference in the remission rates between the early and delayed cohort was more prominent in bio-naive patients. This subpopulation is of special interest because it is exempt from the confounding effect of prior exposure to biologics so that the clinical benefit derived from the earlier initiation of ADL is mirrored closest to its net effect. The greater ADL effect in bio-naive patients is in accordance with that of the SONIC trial, in which the noted benefit of infliximab was at least in part attributed to the inclusion in the study of immunosuppressive and bio-naive patients with a short disease duration (mean of 2.3 years). 33 Beyond bio-naive patients, significant differences between the early and the delayed cohorts were also observed in subgroups with and without steroid dependency/refractoriness and those with and without extraintestinal manifestations at treatment initiation. From a clinical perspective, these results suggest that early intervention with ADL would be highly beneficial when targeting bio-naive patients and even more beneficial when there is coexistence of steroid dependency/ refractoriness and/or extraintestinal manifestations at treatment initiation. In any case, the considerable proportion of patients with extraintestinal manifestations underscores the importance of a multidisciplinary approach towards CD management.
The evaluation of the surgical rates and the rates of progression into complicated phenotypes over time are conceptually associated with the therapeutic mindset of treating promptly within the "window of opportunity" to modify the natural course of the disease, in terms of slowing the progression to stricturing/penetrating phenotypes and/or decreasing the need for CD-related surgeries over time. 34 In the present study, the rates of CD progression into stricturing and/or penetrating phenotypes as well as the CD-related intestinal resection rates were numerically, but not statistically significantly, lower in the early vs delayed cohorts. Initiation of anti-TNF treatment ≤2 years since CD diagnosis has been previously found to be associated with a reduced risk of developing bowel strictures. 20 Additionally, in a study of an exclusively Asian population carrying ≥2 PPFs at baseline, a delayed vs early initiation of anti-TNF or immunomodulators was associated with a reduced risk of developing stricturing and penetrating complications. 23 Further relevant evidence has been generated from a recent retrospective study in which earlier initiation of anti-TNF tended to slow the progression of cumulative bowel damage (P = .069), a term not identical, but overlapping with "CD phenotype progression." 21 A greater rate of surgical resection in the late (>2 years of disease duration) compared with the early anti-TNF initiation cohort was reported in a previous retrospective study of similar design. 22 Similarly, the top-down approach (ie, introduction of anti-TNF agents earlier) has been associated with a lower risk for CD-related surgery. 35 The 52-week ADL survival rate in our study exceeded 90% in the early and delayed cohorts. This rate is somewhat higher than the 1-year rate of 81% reported in a previous study with ADL 36 and the 85% rate reported in a study of patients treated with biologic therapy. 37 The higher rate in our study is likely explained by between-study differences in patient baseline characteristics that affect drug retention rates, including sex, prior use of biologics, disease severity, and the presence of PPF among others. 36 Main limitations of our study include its retrospective design and the missing rate of paired assessments with regard to the HBI at week 26 (26.2% among the early cohort and 29.2% among the delayed cohort). Nonetheless, the difference in the 26-week remission rates between the 2 cohorts with LOCF, NRI, and as-observed analysis consistently point towards a greater derived benefit from early ADL initiation (higher rates in the early cohort), albeit being statistically significant only when using LOCF analysis. The lack of statistical significance with as-observed data and with NRI is likely due to the relatively high missing rate of observations (26% in the early cohort and 29% in the delayed cohort). Other sources for this variation, such as the reason for missing observations, may also have contributed, but such information is not available to be appraised. Another limitation stems from the use of 1-tailed tests for appraisal of the difference between the 2 cohorts in the HBI remission and clinical response rates, which precludes the possibility of detecting an effect in the opposite direction. Of note, the alternate hypothesis of interest of our study, ie, that early initiation of ADL would offer a greater benefit in 26-week remission rates compared to patients initiated with ADL ≥2 years following diagnosis, which formed the basis of using 1-sided testing, was formulated based on published data available at the time of study planning, 18 but also based on recommendations for early introduction of anti-TNF therapy in patients with PPFs. 19,26,27 Moreover, results in the present study are primarily based on clinical symptoms, which do not always correlate with disease activity. The study results are also limited by the fact that over the study observation period there was a small number of observed events regarding intestinal surgeries (other than for perianal disease) and progression to a complicated phenotype, resulting in immaturity of the data used in the survival analyses for estimation of the median times to event. Furthermore, the differences between the 2 cohorts in the rates of CD progression into stricturing and/or penetrating phenotypes as well as the CD-related intestinal resection rates did not reach statistical significance, maybe also due to the overall low number of observed events. The low rate of history of CD-related resections prior to ADL initiation, must be appraised taking into consideration the fact that patients with a history of intestinal resection were excluded from enrollment and that minor surgical procedures, such as seton placement, may not have been consistently recorded in patients' medical charts. It is noted that the main clinical characteristics of the patients did not statistically significantly differ between the 2 cohorts at baseline (Figure 3), thus limiting potential confounding of the study outcomes. It should be acknowledged that 37.1% and 34.9% of patients from the early and delayed ADL cohorts, respectively, already had stricturing/penetrating phenotype at baseline. Although the difference between the 2 cohorts is not statistically significant, these proportions are considerably high, particularly in the early cohort, and may have impacted the results of the study. Furthermore, while the study only examined the impact of early initiation of ADL on disease outcomes in patients with PPFs, the beneficial effects seen in this study may likely be applicable to early initiation of any anti-TNF, which is supported by the more pronounced difference between the 2 cohorts in the bionaïve populations and is aligned with recommendations for early biologic therapy initiation in patients with CD and PPFs. 19,26,27 The major strength of the study is the use of PPFs when investigating the clinical effect of intervening early in the course of CD. To our perception, prognosis is the tool to individualize the definition of early CD, which can be used as a measure of the length of the patient's individual "window of opportunity" to intervene and maximize treatment outcomes. We have attempted to use prognosis to tailor a custom-made treatment paradigm for ADL by articulating the patient profile that would benefit most from early ADL initiation. Conceptually, we believe that the clinical implementation of the PPF would protect patients with poor prognosis from being undertreated and patients with favorable prognosis, whose "window of opportunity" may be as long as their lifetime, from being overtreated. However, to gain the most from the clinical implementation of prognosis, the quantitative validation of the prognostic factors should first be determined through relevant studies. Regarding the generalizability of the study results, it is noted that the study population was enrolled from 10 hospital sites located in diverse geographic locations, aiding representativeness and allowing the reflection of variations in medical practice paradigms.

Conclusion
This real-world, nationwide study emphasizes the short-and medium-term benefits of early vs delayed treatment with ADL in patients with CD who carry a number of PPFs for a long-term disabling disease course. The beneficial effect of early ADL treatment seems to be more pronounced in bio-naive patients for whom treatment has failed or who are dependent on treatment with steroids and experience the devastating effect of severe extraintestinal manifestations. However, well-designed, prospective, real-world studies are needed to confirm our results.

Supplementary Data
Supplementary data is available at Crohn's and Colitis 360 online.

Funding
This work was supported by AbbVie Pharmaceuticals, S.A., Greece. The design, study conduct, and financial support for the study were provided by AbbVie. AbbVie participated in the interpretation of data, review, and approval of the manuscript. All authors had access to relevant data and participated in the drafting, review, and approval of this publication. No honoraria or payments were made for authorship.

Data Availability
AbbVie is committed to responsible data sharing regarding the clinical trials we sponsor. This includes access to anonymized, individual-and trial-level data (analysis datasets), as well as other information (eg, protocols and clinical study reports), as long as the trials are not part of an ongoing or planned regulatory submission. This includes requests for clinical trial data for unlicensed products and indications. These clinical trial data can be requested by any qualified researchers who engage in rigorous, independent scientific research, and will be provided following review and approval of a research proposal and Statistical Analysis Plan (SAP) and execution of a Data Sharing Agreement (DSA). Data requests can be submitted at any time and the data will be accessible for 12 months, with possible extensions considered. For more information on the process or to submit a request, visit the following link: https://www.abbvie.com/our-science/clinicaltrials/clinical-trials-data-and-information-sharing/data-andinformation-sharing-with-qualified-researchers.html.