Association between Endoscopist Specialty and Colonoscopy Quality: Protocol for a Systematic Review and Meta-analysis

Colonoscopy is an important modality in the provision of colorectal cancer screening. Though effective and safe, the performance of screening-related colonoscopy is variable in terms of its overall quality, with endoscopist-related factors playing an important role in this variation. The purpose of this study is to systematically review the association between endoscopist specialty and colonoscopy quality and outcomes. A comprehensive electronic search will be carried out to determine the association between endoscopist specialty and colonoscopy quality metrics and/or outcomes in adult patients undergoing colonoscopy. Two abstracters will independently determine study eligibility, assess study quality, and abstract study data. The primary outcome will be the adenoma detection rate (ADR); secondary outcomes will include cecal intubation rate (CIR), withdrawal time (WT), adverse events (AEs) and post-colonoscopy colorectal cancer (PCCRC). Rates of the above metrics and outcomes will also be compared between clinically relevant subgroups determined a priori. DerSimonian and Laird models will be used to perform meta-analyses for each outcome. Sources of heterogeneity will be explored via meta-regression analyses, if possible. Our


Discussion
Given the signi cant established variation in colonoscopy quality, endoscopist-related factors need to be explored. Our meta-analysis will address the important question of whether the specialty of the endoscopist impacts colonoscopy quality and/or outcomes. Systematic Review Registration PROSPERO CRD42021226251.

Background
Colorectal cancer (CRC) is a leading cause of cancer-related mortality in the United States, where approximately 150,000 associated new cases and 50,000 associated deaths are estimated in 2020 [1].
Screening is effective in reducing the incidence of and mortality from CRC [2]. Screening-related colonoscopy plays an important role in both opportunistic and programmatic screening, being capable of both detection and resection of pre-malignant adenomatous polyps [3]. Colonoscopy is also the recommended procedure following positive primary CRC screening results from other modalities.
Unfortunately, there is widespread variation in colonoscopy quality, with a large proportion of this variation explained by both modi able and non-modi able endoscopist-related factors [4,5].
Several quality indicators have been proposed and are currently widely supported in order to improve colonoscopy quality and reduce variations in care. These include the adenoma detection rate (ADR) and cecal intubation rate (CIR) [6,7]. While generally considered a safe procedure overall, colonoscopy is also associated with rare but potentially serious adverse events (AEs), including bleeding and perforation [8].
Therefore, an additional aim of these quality indicators is to minimize the risk of AEs.
Endoscopist-related factors affect the quality and outcomes of screening-related colonoscopy. In a recent meta-analysis, we evaluated the relationship between endoscopists' annual procedural volumes and colonoscopy quality and outcomes [9]. We reported that higher annual colonoscopy volumes were associated with higher CIR, in addition to observing suggestive trends toward lower rates of adverse events [9]. However, no clear associations were found between colonoscopy volumes and ADR [9]. Thus, other potentially relevant endoscopist-related factors are likely present, the consideration of which could help further elucidate variation in colonoscopy quality.
The specialty of the colonoscopist has been proposed as a potential endoscopist-related factor that could explain variations in quality [10]. Endoscopist specialty intuitively makes sense as a potential mechanism for variation, as many have previously highlighted and questioned differences in formal colonoscopy training between specialties [11][12][13]. Some studies have demonstrated associations between endoscopist specialty and the development of PCCRC [14,15]. However, no synthesized evidence is available on this subject. Thus, we propose a systematic review and meta-analysis in order to elucidate whether the specialty of the endoscopist is associated with either colonoscopy quality indicators or procedural outcomes.

Overview and Objectives
Our study will be carried out according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines [16,17]. Our review is registered on PROSPERO (CRD42021226251).
The overall objective of this review is to determine whether there is an association between endoscopist specialty and colonoscopy quality metrics and/or outcomes when performed in adult patients. The primary objective is to determine whether there is an association between endoscopist specialty and ADR. The secondary objectives will be to determine whether this association exists for any other established colonoscopy quality metric (including withdrawal time [WT] or CIR) or for any colonoscopy adverse event (including bleeding, perforation, or post-colonoscopy colorectal cancer rate [PCCRC]). No research ethics approval is required for this study given there are no patient-level data being collected.

Search Strategy
A detailed electronic search strategy has been designed by the study authors in consultation with a dedicated health research librarian. The electronic databases MEDLINE (Ovid), PubMed, CINAHL, EMBASE, Scopus, Web of Science, and Evidence Based Medicine (EBM) Reviews will be queried from inception of each database to the search date(s). An unabridged search strategy is provided in the Supplementary Materials.

Eligibility Criteria
A study will be included for analysis if it meets ALL of the following inclusion criteria: 1. it is an observational study (prospective or retrospective) or clinical trial; 2. it is written in English; 3. it was published in or after the year 2000; 4. it was conducted in adult patients ≥ 18 years of age undergoing colonoscopy; 5. it reports a direct comparison between at least two separate endoscopist specialties; and 6. it reports one or more of the following colonoscopy quality indicators or outcomes: ADR, CIR, WT, overall or speci c AEs, or PCCRC.
A study will be excluded if it meets ANY of the following exclusion criteria: 1. it reports on shared data (on a patient level or endoscopist level) from another study -in these instances, the largest study will be included; 2. it does not report endoscopist specialty data; 3. it reports trainee performance data (from fellows, residents or medical students) in a non-strati ed cohort; or 4. it reports data in a non-strati ed cohort comprised of upper endoscopic procedures, exible sigmoidoscopy, or pancreatico-biliary endoscopic procedures.
The eligibility criteria are summarized in Table 1. • study reports on trainee performance data in a non-strati ed cohort • study reports on endoscopic procedures other than colonoscopy in a non-strati ed cohort Study Selection and Data Abstraction All citations will be imported into Rayyan (M Ouzzani, Qatar Computing Research Institute, HBKU, Doha, Qatar) and duplicate citations will be consolidated. Two reviewers (MM, AM) will independently screen all captured titles and abstracts in parallel to identify citations for the subsequent stage (full-text review). All full texts will then be reviewed in parallel by the same reviewers. Any con icts from the initial and full-text screens will be resolved by a third investigator (NF).
A standardized data abstraction form will be created in order to capture pre-speci ed data from each study included after the full-text review stage. This form will include relevant study characteristics, patient demographics, endoscopist characteristics, and outcomes of interest. Study-speci c de nitions of outcomes will also be captured. Endoscopist specialty will be carefully designated as accurately as possible as described by individual studies, and these de nitions will be noted. Potential groups will include gastroenterologists, surgeons (including overall, colorectal, and general surgeons), internists, general practitioners, or other. The abstraction process will be carried out in parallel by two independent reviewers (MM, AM) who will then verify one another's work. Any con icts in data abstraction will again be resolved by a third investigator (NF). In the event that a study meets eligibility criteria but lacks data required for meta-analysis, two separate attempts (one week apart) will be made to contact corresponding study authors to obtain any missing data.

Risk of Bias
Two authors (MM, AM) will conduct risk of bias assessments in parallel for all studies included in the nal analyses. Assessment of observational studies will be assessed using the ROBINS-I tool [18], with con icts resolved by consensus with a third investigator.

Statistical Analyses
We will conduct several DerSimonian and Laird meta-analyses to estimate the pooled incidence rates of individual quality metrics and AEs along with 95% con dence intervals (CIs). Study data will be included only if outcomes are clearly de ned and if outcome de nitions are consistent with those used in our previous study assessing the relationship between procedural volume and colonoscopy outcomes [9]. Incidence rates from observational studies and randomized trials will be pooled separately (at no point being combined). To address potential causes of inter-study heterogeneity, we will perform several subgroup analyses in clinically and methodologically relevant groups determined a priori. These are summarized in Table 2. Sources of heterogeneity will also be tested by performing meta-regression. Subgroup analyses and meta-regression will only be performed if there are a minimum of 4 studies in each group [19]. We will also conduct sensitivity analyses where studies of varying quality as per the ROBINS-I tool will be considered separately. Inter-study heterogeneity will be assessed by calculating and reporting the Cochrane I 2 statistic. Publication bias will be assessed by visual inspection of funnel plots as well as Egger's and Begg's tests [20,21]. Revman 5.1 (Cochrane Collaboration) and Stata 14.0 (StataCorp) will be the software packages used for statistical analyses. The overall strength of the body of evidence for each outcome will be assessed using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) framework [22].

Discussion
This systematic review and meta-analysis will provide the rst evidence synthesis on whether the specialty of the endoscopist is associated with colonoscopy quality metrics and/or outcomes. Given the established inter-endoscopist quality variations seen with colonoscopy, the results of this study are likely to guide aspects of training, maintenance of certi cation, credentialing, and quality assurance going forward.
The eld of colonoscopy quality is relatively well established, with a number of gastroenterology societies world-wide having endorsed minimum and aspirational benchmarks for several quality indicators and outcomes [23,24]. These benchmarks were designed based on available evidence at the time and are intended to reduce endoscopist-level variation and consequently improve patient outcomes. However, signi cant variation persists in the quality of colonoscopy, an integral procedure in the overall effort to screen for and reduce the burden of CRC [3]. Thus, our study represents an important piece of the puzzle in terms of determining potential endoscopist level factors that are associated with quality metrics and outcomes.
Though our study protocol was designed to mitigate potential biases, our study will nevertheless have limitations. Firstly, given our experience performing meta-analysis on endoscopist factors in colonoscopy [9], we anticipate that several studies will not report potentially relevant confounding variables such as endoscopist age, cumulative experience, procedural volumes, procedural indications, doses of sedating medication, or bowel preparation scores. Furthermore, even if studies do report potential confounders, it is anticipated that most studies assessing our topic of interest will report these data at the study level (rather than at the endoscopist level). Thus, the results of any potential subgroup analyses or metaregression analyses could be misleading given the potential for ecological bias [25]. Additionally, there may be an element of selection bias within each study, whereby surgeons could conceivably have different case mixes than gastroenterologists; this could conceivably bias the results in favour of one group or another when one considers quality indicators such as ADR or adverse event outcomes such as perforation. To mitigate against this risk, we have set strict a priori criteria that de ne included study samples as primarily screening populations versus non-screening populations ( Table 2).
In a broader context, we also anticipate that the relevance and actionability of our study ndings will be questioned. Especially when one considers the shift toward competency-based medical education in general [26], and in endoscopic training in particular [27], the current comparisons between gastroenterologists, surgeons, and other endoscopists may become less relevant over time. However, we maintain that this topic is of urgent importance to the current landscape as it pertains to colonoscopy practice. Nevertheless, the practice of audit and feedback of quality metrics as well as educational interventions for independent colonoscopy practitioners are more relevant now than ever, having both been shown to be effective in improving colonoscopy quality [28,29].
Overall, despite these limitations, we anticipate that our study will yield important information on endoscopist specialty and its association with colonoscopy quality indicators and outcomes. Our results have potential implications regarding not only direct patient care, but credentialing and training. Our study will summarize the existing state of evidence on this topic, and in so doing, will also serve to elucidate potential improvements for future studies assessing endoscopist level characteristics in the performance of colonoscopy. Authors' contributions NF conceived of and designed the study, co-developed the analysis plan, and drafted the article. DRB contributed to study design and co-developed the analysis plan. YR and MM co-developed the analysis plan. SJH helped design the study. All authors are responsible for critical revision of the article for intellectual content, and nal approval of this manuscript.