Considerations for Colorectal Neoplasia Detection in Inflammatory Bowel Disease Clinical Trials

Colorectal cancer (CRC) is the third most common cause of cancer-related mortality worldwide. Patients with inflammatory bowel disease (IBD) are at higher risk for developing CRC compared with the general population, and CRC accounts for 10–15% of all-cause mortality in patients with IBD [1, 2]. The risk of CRC generally begins to increase approximately 8–10 years after symptom onset and increases linearly over time [3]. In population-based cohorts, patients with ulcerative colitis (UC) or extensive colonic Crohn’s disease have a 2- to 3-fold increased risk of CRC [2, 4‒8]. Other factors associated with the magnitude of additional risk include male sex; younger age at diagnosis; concurrent primary sclerosing cholangitis (PSC); personal or family history of colorectal dysplasia; and extent, severity, and duration of inflammation, defined endoscopically and histologically [1, 2, 8‒11].

Endoscopic surveillance in patients with IBD is associated with a higher rate of CRC detection, including for early-stage CRC, as well as a lower rate of CRC-associated deaths when compared to no surveillance [12, 13]. Early detection is essential for improving CRC-related prognosis and is dependent on patient-related factors (e.g., bowel preparation quality), high-quality endoscopy, and accurate histologic diagnosis (Fig. 1; Table 1) [14]. Surveillance colonoscopy for patients with IBD at increased risk for CRC has been endorsed by multiple societies, although frequency of surveillance and recommended modalities for evaluation vary [4, 11, 15‒18]. Guidelines from the European Crohn’s and Colitis Organization (ECCO) recommend colonoscopy every year for high-risk patients (stricture or dysplasia detected within the past 5 years, PSC, extensive colitis with severe active inflammation); every 2–3 years for intermediate-risk patients (extensive mild or moderate active inflammation, postinflammatory polyps, or a family history of CRC in a first-degree relative diagnosed age 50 years or older); and every 5 years for low-risk patients (neither intermediate nor high-risk features) [19]. Risk level is dependent on factors including the cumulative burden of endoscopic and histologic inflammation, concomitant PSC, anatomic abnormalities (e.g., lead pipe or shortened colon), and a personal or family history of CRC [4, 11, 19‒21].

IBD patients who are considering participation in clinical trials may have clinical features associated with increased risk for colorectal neoplasia (CRN), which includes both CRC and dysplasia, due to factors such as active mucosal inflammation and long-standing disease duration. Generally, the risk of CRC is thought to be reduced when mucosal inflammation is better controlled with potent immunosuppressive therapies; however, some immunomodulatory investigational products may be associated with an increased risk of malignancy by attenuating cancer immunosurveillance [22]. Consequently, clinical trial eligibility criteria are designed to exclude patients who have the highest risk for developing malignancies at baseline as well as those with CRN at the baseline screening endoscopy. Conversely, patients with IBD enrolled in clinical trials are often monitored more closely than they are in routine practice; given the increasingly important role of endoscopy as a measure of treatment safety and efficacy, patients typically undergo more intensive endoscopic evaluation in a trial setting. Effective treatment of mucosal inflammation, a desired outcome in IBD clinical trials, is associated with a decreased risk of IBD-associated CRN [10, 23‒25]. The objectives of this narrative review were to: (1) highlight key considerations of CRN detection within IBD clinical trials and to discuss evidence-based best practices; (2) consider current CRN identification practices within IBD clinical trials from the lens of the sponsor, site endoscopist, and central read team; and (3) provide suggestions to improve overall surveillance and detection of CRN.

Clinical trials must be carefully designed to define eligible populations, use reliable tools to measure disease activity, and provide comprehensive outcome measures for both efficacy and safety [26]. IBD clinical trials must consider factors such as type and severity of disease, inclusion based on lack of response or intolerance to previous treatment(s), concomitant or previous use of medications, medical history, frequency of study visits, and frequency and extent of endoscopic procedures. Historically, stringent eligibility criteria resulted in limited generalizability of clinical trial findings, as less than one-third of patients with IBD seen in clinical practice were eligible for trial enrollment [27]. Patients with a history of malignancy were often excluded from trials, as the risk profiles of therapies under investigation for IBD are often not completely known. However, there is a need to enroll a population more reflective of the “real world” that will be exposed to the drug once it is marketed. Registrational trials are generally powered for the primary efficacy endpoint and are underpowered to detect infrequent adverse events such as dysplasia or malignancy; to address the exclusion of patients with comorbidities, long-term prospective post-marketing registries are essential for establishing a comprehensive safety profile [28].

Table 2 outlines the exclusion criteria for five phase 3 clinical trials based on medical history and screening, screening procedure, number of endoscopies, and incidence of CRC [29‒35]. All trials excluded patients with IBD based on history or findings of malignancy at screening, with the exception of treated nonmelanoma skin cancer and cervical carcinoma in situ. Protocols for golimumab, tofacitinib, and vedolizumab included patients with a history of adenomatous polyps contingent on complete removal. The instrument of choice for endoscopy varied across clinical trials. Protocols for adalimumab and golimumab in UC did not specify the instrument or extent of endoscopy, while the protocol for tofacitinib required a full colonoscopy prior to baseline for high-risk patients (Table 2). Flexible sigmoidoscopy for UC is less time-consuming but still assesses disease activity and may detect other conditions such as Cytomegalovirus or Clostridioides difficile infections; however, it examines only the distal colon and may lead to undetected proximal inflammation and dysplasia [36‒38]. Hence, more sponsors are recommending a full colonoscopy over sigmoidoscopy at screening, if one had not been performed in the preceding year [39, 40]. Sponsors determine total number of endoscopies during induction and maintenance phases. Aside from the initial screening endoscopy, generally one additional endoscopy is performed to assess induction response (typically at weeks 8–12) and another is performed during the maintenance phase (Table 2). No research exists differentiating endoscopy frequencies during the course of a clinical trial with regard to CRN detection. Given that CRN is a serious but infrequent adverse event, endoscopy frequency is typically based on what is needed to determine efficacy of treatment.

It is recommended that sponsors set minimum standards for site colonoscopy acquisition times and bowel preparation as well as central reader viewing times [41]. The US Food and Drug Administration (FDA) Clinical Trial Imaging Endpoint Process Standards recommend that logistic and technical considerations such as image modality, timing and frequency of image evaluation, image quality control, and use of a central reader remain with the sponsor. While there are no guidelines applicable across all phase 3 clinical trials, sponsors should work in accordance with national and local scientific guidelines alongside central reading vendors to standardize imaging procedures through an imaging charter [42]. In addition, a communication plan for reporting incidental findings should be outlined. The sponsor should address incidental findings in the informed consent forms, which are reviewed and signed at study entry, and have a well-developed plan to address CRN discovery [43]. Guidelines for evaluation, verification reporting, and follow-up of incidental findings should be included [43, 44].

Finally, the emerging importance of histologic improvement for clinical outcomes has led all ongoing IBD intervention studies to include mucosal biopsy as part of trial procedures [45]. Sponsors should consider real-time histopathology analysis of biopsy samples, rather than batch analyses, because of possible incidental discovery of dysplasia and implications of delayed diagnosis. Overall, the sponsor plays a crucial role in CRN surveillance through its role in selecting the patient profile, determining the study and procedure design, and enforcing study conduct.

The site endoscopist serves multiple crucial roles in IBD clinical trials. The endoscopist operationalizes the protocol’s endoscopic requirements, either functioning as or working closely with the principal investigator to implement protocol procedures used in conjunction with individual patient needs. The screening procedure and instrument used are decided by the endoscopists, as outlined in Table 2. Hence, endoscopists should obtain a comprehensive risk profile based on patient characteristics, in accordance with the investigator guidance document. To accurately assess disease activity and optimize opportunities for CRN detection, the site endoscopist must also perform high-quality endoscopy and record optimal video for the central reader. Videos of suboptimal or unreadable quality may lead to inaccurate interpretations by central readers, increasing chances of discrepant reads between site endoscopists and central readers [41]. Furthermore, the site endoscopist is responsible for decisions that occur during endoscopy, including where, when, and how long to look for incidental findings and how to address suspicious lesions. Hence, a communication plan for reporting and follow-up of incidental findings should center on the site endoscopist.

Currently, no validated instruments or consensus agreements for measuring endoscopy quality exist for IBD clinical trials [46]. However, there are rational best practices that endoscopists should consider, including video-related and procedural factors. Video-related metrics include using high-definition colonoscopes, recording high-resolution videos, and capturing accurate colors, shapes, and textures [47]. Procedural factors include cleaning, washing, and suctioning any residual fecal debris, visualizing the entire lumen, minimizing reflections/glares with proper endoscope positioning, and withdrawing at a sufficient pace to completely visualize the mucosa.

As the operator of the colonoscope, the site endoscopist is primarily responsible for the careful and comprehensive evaluation of the colonic mucosa and appropriate acquisition of endoscopic biopsies. It should be noted that in the context of active inflammation, it is often challenging (if not impossible) to accurately detect and diagnose dysplastic lesions. This is typically the case at screening for a clinical trial, where patients require a minimum burden of active endoscopic inflammation to qualify for participation. Histologically, it can also be difficult to distinguish changes arising from inflammation from those related to true dysplasia. For example, epithelial regeneration and repair in the setting of active inflammation can result in histologic atypia that may be interpreted as “indefinite for dysplasia” [48]. This highlights that both the endoscopist and pathologist must be diligent in evaluating the mucosa carefully for potential dysplastic changes, especially when these findings may be subtle or masked by inflammation.

In 2021, ECCO released guidelines around quality reporting of colonoscopy in patients with IBD, many of which can be extrapolated to clinical trials [49, 50]. Pre-procedure, sites should ensure that patients are counseled on the importance of and specific instructions for bowel preparation; furthermore, patients should be adequately sedated to ensure procedure tolerability, and endoscope and processor should be set up appropriately for trial video capture. Intra-procedure, digital anorectal examination should be mandatory to ensure that physical examination findings are documented for the trial, rectal retroflexion is considered, and colon and ileum are explored.

In clinical practice, colonoscopy quality is indicated primarily by surrogate measure of adenoma detection rate (ADR) [51, 52]. ECCO has identified effective bowel preparation, slow colonoscopic withdrawal, and use of high-resolution endoscopic equipment as preferred tools for optimizing detection of neoplasia, all of which influence ADR [19, 49]. Finally, patient satisfaction and compliance should also be considered as measures of surveillance success.

The American College of Gastroenterology and the American Society for Gastrointestinal Endoscopy Task Force on Quality recommend a withdrawal phase of at least 6 min for colonoscopy from the terminal ileum to the rectum in patients without IBD [52]. European quality initiatives in 2019 have established a minimum standard time of 6 min with a target mean standard time of 10 min [53]. A recent multicenter, randomized, controlled trial including 1,027 patients has demonstrated that prolonging withdrawal times from 6 to 9 min significantly increased ADR from 27.1% to 36.6% (p < 0.001) [54]. In a clinical trial, withdrawal times should be even longer for patients with IBD to allow for careful examination for CRN and to ensure that adequate washing and proper visualization for central reading is accomplished. Keeping records of withdrawal times can improve ADR and accountability; these measures can be monitored using anatomical landmarks recorded for central reading [55, 56].

Withdrawal technique is another important indicator of colonoscopy quality; poor withdrawal technique as assessed by inadequate fold examination, distension, and cleansing of five areas of the colon is associated with low ADRs [57, 58]. Adequate bowel preparation is essential for adenoma and polyp detection; it should be palatable and well-tolerated while rapidly and reliably clearing the colon of all fecal material, leaving no gross or histologic alterations in the mucosa [50, 59‒61]. Recent FDA guidance indicates that an excellent bowel preparation should include only small bits of fecal residue or fluid, requiring no washing or no/minimal suctioning with an overall clear visualization of the entire colonic mucosa [62].

Inadequate bowel preparation can lead to prolonged or incomplete procedures, increased costs due to multiple procedures, missed lesions, and higher complication rates [59, 63, 64]. For patients with IBD who invariably require multiple colonoscopies throughout their lifetime, the tolerability and acceptability of bowel preparation are important considerations [65, 66]. Patients with IBD consistently report bowel preparation as one of the most burdensome aspects of the procedure, and this negatively impacts adherence to colonoscopy surveillance recommendations [65‒67]. Furthermore, actively symptomatic patients with abdominal pain, nausea, or vomiting may be even less tolerant of large-volume liquid bowel preparations for colon cleansing. Recently, the FDA approved an oral tablet bowel preparation (sodium sulfate, magnesium sulfate, potassium chloride) as an alternative to liquid-based osmotic laxatives [68]. Therefore, advances aimed at improving bowel preparation tolerability hold promise for patients with IBD. It should be cautioned that some contact stimulant laxatives such as sodium picosulfate have been associated with colonic and ileal erosions that may confound interpretation of disease activity in patients with IBD.

In clinical practice, high-definition white-light endoscopy with random biopsies and chromoendoscopy with targeted biopsies have been identified as appropriate methods for dysplasia surveillance in patients with IBD [18]. An analysis of 10 prospective studies showed that, on average, only one focus of dysplasia was detected for every 1,505 random biopsy specimens taken, evidence that random biopsy is a time-consuming and cost-inefficient approach on its own [69]. However, there may be scenarios where random biopsies are still clinically useful. For example, endoscopic surveillance or chromoendoscopy may be difficult to perform in patients who have persistent inflammation or who have inadequate bowel preparation. In these settings, random biopsies are frequently taken. Furthermore, some clinicians also take random biopsies in addition to targeted biopsies for those patients at highest risk, including those with PSC, previous dysplasia, and/or tubular or foreshortened colon [70].

Alternatively, chromoendoscopy with targeted biopsies has been shown to improve dysplasia detection over standard or high-definition white-light colonoscopy with random biopsies [71‒73]. Hence, recent recommendations favor the use of chromoendoscopy or virtual chromoendoscopy with technologies such as narrow-band imaging or digital optical enhancement with targeted biopsies [19, 27, 74]. For white-light endoscopy, random and targeted biopsies must be performed, and random biopsies are still recommended for patients with unique circumstances such as concomitant PSC, a foreshortened colon, or history of dysplasia [4, 15, 75‒77]. The role of chromoendoscopy in the clinical trial setting is less clear; it may be challenging to discern certain lesions after dye spray application for enhancement of dysplasia surveillance, and not all sites may perform chromoendoscopy during routine procedures. Therefore, it is recommended that in the clinical trial setting, each colonic segment should first be visualized under white-light examination to delineate the disease activity. Subsequently, if chromoendoscopy is indicated for dysplasia detection, this can be done in a second pass. Additional time should be booked for trial procedures to ensure that adequate visualization both under white light and dye spray can be assessed. It should be noted that the evidence supporting the use of dye spray or virtual chromoendoscopy is primarily based on detection of dysplasia in the setting of endoscopic remission. This is typically not the case at screening, nor would it apply to patients with persistent endoscopic inflammation at follow-up.

If dysplasia is detected, careful endoscopic assessment is critical for determining the appropriate next steps in management [74]. It is recommended that lesions are described using the modified Paris Classification, as either polypoid (≥2.5 mm tall, pedunculated or sessile), nonpolypoid (<2.5 mm, elevated, flat, or depressed), or invisible (detected on random biopsy). Previously used terms including dysplasia-associated lesion or mass (DALM), adenoma-like mass, or flat dysplasia are ambiguous and should not be used. All suspicious lesions should be assessed for endoscopic resection. A “five S” approach has been proposed for evaluating the shape, size, site, surface Kudo pit pattern, and surrounding evidence of mucosal inflammation or other lesions [49]. Generally, lesions <2 cm with a clearly demarcated border, no features of submucosal invasion or fibrosis, and no histologic features of invasive carcinoma should undergo an attempt at endoscopic resection. Specialized expertise in therapeutic endoscopy should be sought for larger lesions, more complex laterally spreading or irregular lesions, or lesions with local recurrence or incomplete resection after previous attempts. A multidisciplinary approach with surgical colleagues is recommended for such complex cases.

The overall tenets of high-quality endoscopy for site endoscopists have recently been summarized in the 2022 European Society of Gastrointestinal Endoscopy (ESGE) performance measures for colonoscopy in patients with IBD [46]. The authors agreed that indications for colonoscopy in IBD (including endoscopic assessment of disease activity and surveillance) should be explicitly documented and that disease activity should be described using standardized scoring systems, such as the Mayo Endoscopic Subscore or Ulcerative Colitis Endoscopic Index of Severity for UC, or the Crohn’s Disease Endoscopic Index of Severity or Simple Endoscopic Score for Crohn’s Disease [78]. There was also agreement that at least 90% of colonoscopies performed in IBD patients should have adequate bowel preparation, high definition should be universally used for surveillance in patients with long-standing colitis, and that either dye-based or virtual chromoendoscopy in combination with targeted biopsies should be used in surveillance with a minimum standard of this being achieved in ≥70% of cases. The minimum detection rate of neoplasia in surveillance colonoscopies was recommended to be ≥8%.

The site endoscopist is optimally positioned for CRN surveillance with the highest-resolution image, knowledge of a patient’s history and risk profile, and facets of the endoscopy procedure itself, including the scope technique, biopsies, and resection of lesions. However, the central read team can provide a secondary layer of protection for CRN detection, recognizing that the central reader assessment depends on the submitted video quality [79, 80]. A challenge faced by central readers is when and how to report incidental findings, such as potentially dysplastic polyps, which often have already been evaluated with better resolution by the site endoscopist. In some instances, inflammatory or hyperplastic polyps may be visualized based on endoscopic appearance, and further endoscopy may not be needed based on the site endoscopist’s assessment and judgment. In these situations, it would seem appropriate for the central reader to not report these as incidental findings; however, the distinguishing endoscopic features of benign polyps are often not seen from a central reader perspective. In addition, if there is a concern that a polypoid lesion is dysplastic, it would be prudent for the central reader to report the finding, and the reporting process should allow for this provision.

The central read workflow involves collaboration between many individuals, including image services coordinators who confirm and communicate video assessment, subject eligibility, and video quality; project coordinators who note trends in suboptimal videos; clinical trial managers who assess site performance; vendor medical team that performs supplementary video review; and central readers on a Code of Federal Regulations 21 part 11-compliant read system. The results from the assessment, including any incidental findings or quality issues, should then be communicated and followed with the site endoscopist and/or principal investigator. This multistep workflow and oversight must be an efficient process with few technical challenges and allow for modification as needed.

To adhere to CRN detection and surveillance standards set by the sponsor, central readers should be expert endoscopists experienced in the assessment of disease activity and detection of endoscopic dysplasia and must complete assessment of training video cases per the independent review charter [41]. At minimum, baseline qualifications include board certification in gastroenterology and experience and post-certification training in managing patients with IBD [41, 50]. Central reader performance metrics should be benchmarked at onboarding, and then subsequently maintained on at least annual assessments. Generally, central readers should demonstrate at least substantial inter- and intrarater agreement (as measured by an intraclass correlation coefficient >0.60 per the Landis and Koch benchmarks) [81]. Central readers must also remain adaptive to video/image issues, video quality reports, protocol amendments, and follow-up requests, among other informative notifications [41, 50]. Inter-reader and intra-reader variability are concerns for video/image assessment with both local and central reading practices. Variability can potentially be reduced through standardization of training programs, examination procedures, and reference materials [41, 50, 80, 82]. A central reader should also maintain continuous communication with the sponsor and other readers to provide feedback, discuss video review, and improve read criteria. Balancing turnaround time and prompt image interpretation for quality control with thorough analysis should also be considered; central readers should be prompt when identifying technical flaws and suboptimal videos but should follow minimum established review times [42].

Finally, following a procedure outlined in trial documents, such as investigator guidance for the incidental detection of findings suspicious for CRN, is critical. Central review-directed follow-up and requirements for additional investigations or surveillance after a cancer finding are also suggested. This may be done by consulting a medical team, often through a contract research organization medical monitor, provided by the sponsor to coordinate appropriate follow-up with the site investigator [83]. While the threshold for identifying incidental findings varies by observer, it is recognized that aggressively “over-calling” endoscopic findings (e.g., subtle lesions that are not definitively dysplastic) may result in additional burden for sites and patients. Nevertheless, the potential consequences of missing dysplasia in this high-risk population can be substantial; therefore, a lower threshold for identifying dysplastic incidental findings may be appropriate [43]. The process for addressing incidental findings requires close communication between the sponsor, site investigator, central reader, and patient (Fig. 2) [43].

Best practices for CRC identification in IBD clinical trials require engagement and collaboration between the clinical trial sponsor, the site endoscopist and/or principal investigator, and the central read team (Table 3). In the protocol, the sponsor should be explicit with eligibility criteria regarding a history of malignancy and should outline follow-up procedures if suspicious lesions, dysplasia, or malignancies are found at the initial screening endoscopy. Furthermore, sponsors should ensure that family history, risk factors for IBD, and other personal risk factors for CRC are considered when generating comprehensive risk profiles, and that a record of recent colonoscopies from all patients is available. Sponsors should require patients to undergo colonoscopy if their most recent procedure was over 1 year ago. The instruments and appropriate biopsy procedures can be specified in the protocol, as chosen by endoscopists based on a patient’s risk profile; patients who are at high risk should have a full colonoscopy rather than a flexible sigmoidoscopy.

The sponsor should prepare a site guidance document specific for the protocol that should serve to set minimum standards for site endoscopists and central readers based on evidence-based best practices. Site endoscopists should focus on slow, proper withdrawal technique, and adequate mucosal visualization of the terminal ileum to rectum. Adequate bowel preparation, proper labeling of bowel segments, and adequate washing are also essential in addition to focusing on taking optimal color videos using proper lighting for mucosal assessment. To improve image quality, an imaging charter can be developed specifying an optimal image quality as well as outlining file size and bandwidth requirements. Central readers should adhere to standards for image and video review times to balance quick turnaround time with thorough evaluation. In collaboration with imaging laboratory teams, the sponsor should also provide site training and updated educational tools to site endoscopists and central readers to improve performance.

When creating site guidance documents, the sponsor should collaborate with site endoscopists, central readers, and principal investigator to draft a communication plan for disclosing incidental findings and follow-up. Specific considerations may include noting the presence or absence of incidental findings such as lesions or polyps by the site endoscopists, which in turn provides awareness for central read review. As part of the communication plan, central readers should provide site-specific feedback to improve detection performance and video quality of the site endoscopist. Consultation between the vendor medical team and site endoscopist is recommended to provide second opinions and supplemental reviews as needed. Patients should be made aware of these procedures to address incidental findings through informed consent documents.

IBD clinical trials provide an important opportunity for extended endoscopic evaluation and CRN identification. Through understanding key synergistic relationships between team members, evidence-based improvements can be made to further the field and optimize outcomes for patients with IBD in clinical trials.

This study was sponsored by Bristol Myers Squibb. Professional medical writing and editorial assistance were provided by Gorica Malisanovic, MD, PhD, at Peloton Advantage, LLC, an OPEN Health company.

An ethics statement is not applicable because this study is based exclusively on published literature.

M.M.Y. was an employee of Bristol Myers Squibb at the time of manuscript initiation. H.A.A., S.A., and A.S. are employees and shareholders of Bristol Myers Squibb. J.B.C. was an employee of Bristol Myers Squibb at the time of manuscript submission. K.U. was an employee of Bristol Myers Squibb at the time of manuscript initiation and currently serves as a paid consultant. F.A.F. has received consulting fees from Arena, BMS, Braintree Labs, GI Reviewers, GSK, IBD Educational Group, Innovation Pharmaceuticals, Iterative Scopes, Janssen, Pfizer, and Sebela. He serves on a data safety monitoring board for Bacainn Therapeutics, Lilly, and Theravance. C.M. has received consulting fees from AbbVie, Amgen, AVIR Pharma Inc, Ferring, Fresenius Kabi, Janssen, McKesson, Mylan, Takeda, Pfizer, Roche, and Alimentiv (formerly Robarts Clinical Trials Inc.); speaker’s fees from AbbVie, AVIR Pharma Inc, Janssen, Takeda, and Pfizer; and research support from Pfizer.

Professional medical writing and editorial assistance were funded by Bristol Myers Squibb.

Study design: M.M.Y., K.U., C.M., H.A.A., S.A., and J.B.C. Study investigator and Enrolled patients: N/A. N/A. Collection and assembly of data and manuscript preparation: K.U., M.M.Y., H.A.A., S.A., and C.M. Data analysis: M.M.Y., K.U., H.A.A., S.A., A.S., J.B.C., F.A.F., and C.M. Data interpretation, final approval of manuscript, and manuscript review and revisions: M.M.Y., K.U., H.A.A., S.A., A.S., J.B.C., F.A.F., and C.M. The manuscript, including related data, figures (except Fig. 1, being reprinted with permission), and tables, has not been previously published and is not under consideration for publication elsewhere.

Mira M. Yang and Keith Usiskin were employees of Bristol Myers Squibb at the time of the manuscript development.James B. Canavan was employed by Bristol Myers Squibb at the time of submission.

The Bristol Myers Squibb policy on data sharing may be found at https://www.bms.com/researchers-and-partners/independent-research/data-sharing-request-process.html. Further inquiries can be directed to the corresponding author.