A prospective international multisite randomized controlled trial of water exchange with and without distal cap(s) in adenoma detection

Introduction

Data mining studies have reported an association between a low adenoma detection rate (ADR) and interval cancers diagnosed before the next recommended surveillance colonoscopy [1]. New modifications of conventional colonoscopy have been proposed, aimed at decreasing interval cancers by identifying and removing additional premalignant adenomas [2]. A benchmark ADR has been proposed to encourage and guide improvement (e.g., with water exchange added to conventional colonoscopy) [3].

Looking behind folds increases the exposure of the mucosa [4]. A randomized controlled trial in conventional colonoscopy, reported by high detectors in the United States in 2012, showed that a distal straight cap (CAP) increased ADR by 13% (69% vs. 56%, P=0.009) [5]. An earlier Asian report, however, showed a contrary result (30.5% vs. 37.5%, P=0.018) [6], possibly due to residual feces lodged in the CAP interfering with mucosal inspection. No published ADR data existed for the Daisycuff^™ or Endocuff Vision^® in conventional or water exchange colonoscopy at the time of planning of this study protocol (2014-2016).

Adequate bowel cleanliness, coupled with an enhanced inspection technique, improves ADR [7]. Water exchange (the infusion and suction of water during colonoscope insertion) augments the removal of residual fecal debris, resulting in a cleaner colonic lumen for withdrawal inspection [8]. Three in-press randomized controlled trials in 2017 [9-11] showed that water exchange significantly increased ADR when compared with conventional colonoscopy. We postulated that CAP, which facilitates exposure of the mucosa, could increase ADR when added to water exchange.

Unpublished preliminary data suggest that Daisycuff^™ [12] added to water exchange tends to increase ADR. Daisycuff^™ was the initial planned controlled comparison. Insufficient compliance documentation in 2016, however, led to unexpected suspension of approval of Daisycuff^™ by the local institution review boards at 2 of the 7 study sites. The commercially available Endocuff Vision^® [13] was substituted to preserve the participation of coinvestigators committed to performing the comparison between water exchange with and without CAP.

We tested the hypothesis that, in patients undergoing colonoscopy for screening, surveillance, or positive fecal immunochemical test (FIT+) or fecal occult blood test (FOBT+), ADR would be higher for water exchange with CAP than for water exchange with No cap. Daisycuff^™ and Endocuff Vision^® were included as comparative controls.

Patients and methods

This prospective, multicenter, randomized controlled trial was performed at 7 sites (3 in Asia, 1 in Europe, and 3 Veterans Affairs centers in the United States). The diversity of investigators, sites and patient mix aimed to maximize the generalizability of the results. Recruitment took place from July 2018 to October 2022, and was closed during the COVID-19 pandemic. The protocol was approved by Institutional Review Boards at each study site, and by the University of California at Los Angeles. It was registered with ClinicalTrials.gov (NCT03566615). Signed informed consent was obtained.

Management of polyps

To optimize insertion time, all polyps were removed during withdrawal. Small polyps (size <7 mm) were biopsied with forceps. Polyps >7 mm were removed by snare (hot or cold). Very large polyps with features suggestive of malignancy were biopsied for diagnosis, with removal in the same session, or referred to a local interventionalist. All resected polyps underwent pathology assessment by local pathologists.

Fig. 1 shows the distal attachments inside the distended colon during withdrawal. CAP was a transparent straight cap (Disposable Distal Attachment; Olympus Medical Systems Corp., Tokyo, Japan) [5,6,15]. The comparative Daisycuff^™ cap (Visualization Balloons LLC, West Caldwell, NJ) [12] was a ring with 10 “petals” distributed around the circumference. One Daisycuff^™ was placed at the 20-cm mark and 1 at the distal tip (manufacturer’s instruction). The colon was pleated on the instrument shaft. Endocuff Vision^® (Olympus) [13] had a single row of 8 flexible arms. When the tip of the colonoscope was pressed against a fold, an enhanced view of the back side of the fold was obtained.

Figure 1 Distal attachments are depicted in the withdrawal view with distension of the colonic lumen. Straight cap is CAP

Results

A total of 3794 patients expressed interest in participation, and 2280 of them were deemed potentially eligible. Of these patients, 700 declined consent and 200 met the exclusion criteria (Fig. 2): thus, 1380 were consented and examined with water exchange. They were randomized to No cap (N=464), CAP (N=456) and comparative caps (N=460). Randomization distributed the subjects evenly (Supplementary Table 1): mean age 60 years, 30% female, body mass index 26-27 kg/m² and smoking history 36%. The proportions of patients with a family history of colon cancer, specific medical history, self-reported abdominal discomfort, use of narcotic medication, prior colonoscopy, indications of colonoscopy, race and ethnic composition, household income, employment and education status were comparable.

Figure 2 Study flow chart. A total of 3794 participants expressed interest when informed of the study. Of these, 1380 patients consented and were randomized. *Missing data included 3 cases in the CAP group and 8 cases in the combined Daisycuff^TM and Endocuff Vision^® group. FIT, fecal immunochemical test; FOBT, fecal occult blood test; CAP, straight cap

The indications for colonoscopy in the current trial are shown in Table 1. Table 1 also shows some site-related variations in published historical water exchange ADRs, which reflected endoscopist experience, patient mix, procedural-related factors, or local institutional context. In keeping with the original design to assess aggregated data, the exact site identification is not displayed.

Table 1 Site-related variations in historical baseline ADR and indications

Table 2 shows procedural data. Water exchange was applied appropriately during insertion to the cecum with nearly equal mean volumes of water infused 852 (564) mL (n=452) and suctioned 818 (592) mL (n=450) in the No cap group. The other cap groups showed similar patterns. Cecal intubation failed in <5%. Poor bowel preparation was observed during withdrawal in <6%. The withdrawal time did not differ among the groups.

Table 2 Procedure-related characteristics by randomized group

To ensure appropriate comparison by investigators, the groups that randomized similar caps were analyzed separately. Inclusion of all 7 sites maximized the number of patients in the No cap vs. CAP comparison. The ADR was 45.6±2.3% (No cap) and 49.6±2.3% (CAP) (P=0.227) (Table 3A); and CAP significantly increased the ADR in the right colon: 29.7±2.1% vs. 23.5±2%, (P=0.034; 95%CI -12.1% to -0.3%) (Table 3B). Data from the 5 sites that randomized No cap, CAP and Daisycuff^™ showed Daisycuff^™ to be marginally superior in terms of ADR: Daisycuff^™ vs. No cap 52.8±2.8% vs. 44.9±2.8% (P=0.050; 95%CI -16% to 0.3%). Daisycuff^™ significantly improved ADR in the right colon: 32.0±2.7% vs. 24.4±3.4% (P=0.027; 95%CI -15.4% to -0.6%) (Table 3C). In the left colon, Daisycuff^™ had a significantly better adenoma per colonoscopy score (APC): 0.417±0.812 vs. 0.272±0.616 (P=0.012; 95%CI -0.259 to -0.031) (Table 3D).

Tables 3E and 3F show that Endocuff Vision^® did not significantly change ADR or APC.

Table 3A ADR and APC: No cap vs. CAP (data from 7 sites). Data are mean and confidence interval

Table 3B ADR and APC in right vs. left colon – No cap vs. CAP (data from 7 sites)

Table 3C ADR and APC analysis of 5 sites that randomized No cap vs. CAP vs. Daisycuff^™

Table 3D ADR and APC in right vs. left colon – No cap vs. CAP vs. Daisycuff^™ (data from 5 sites)

Table 3E ADR and APC analysis of 2 sites that randomized No cap vs. CAP vs. Endocuff Vision^®

Several ADR differences were nearly significant. For example, overall ADR, Daisycuff^™ (52.8%) vs. No cap (44.9%) (P=0.05; 95%CI -16% to 0.3%) (Table 3C); Endocuff Vision^® 58.0±4% vs. No cap 47.0±4.1% (P=0.057; 95%CI -22.9% to 0.9%) (Table 3E); and Endocuff Vision^® for both right colon ADR 32±3.8% vs. 22.5±3.4% (P=0.066; 95%CI -20.2% to 1.2%) and left colon ADR 47.3±4.1% vs. 35.8±3.9% (P=0.0422; 95%CI -23.3% to 0.2%) (Table 3F).

Table 3F ADR and APC in right vs. left colon – No cap vs. CAP vs. Endocuff Vision^® (data from 2 sites)

Table 4 shows a univariate and multivariate logistic regression analysis of the aggregated data. Long withdrawal time, alcohol history and older age were associated with high adenoma counts (APC≥3) in the multivariate logistic analysis (Table 4B).

Table 4A Univariate logistic regression analysis with adenoma per colonoscopy as dependent variable and individual covariates as independent variables

Table 4B Multivariate logistic regression

Supplementary Table 2A shows site-related variations in study outcome ADR and Supplementary Table 2B shows ADR data by site which did not have significant effects on ADR.

Supplementary Tables 3A-C show ADR comparisons of No cap vs. respective caps based on colonoscopy indication. Only in the case of Endocuff Vision^®, was the ADR of a FIT+ indication significantly higher than that of No cap 66.2% vs. 57.6% (P=0.026; 95%CI -0.372 to -0.012). Supplementary Table 4 shows that varying the mode of sedation did not affect the ADR of different caps, except that CAP had a significantly greater ADR compared with No cap (P=0.008; 95%CI 0.25-0.463) in subjects with no sedation. Supplementary Tables 5-11 show that site, polyp size (large, diminutive, small), shape (sessile, flat, pedunculated) and pathology (serrated) did not affect the ADR of different caps. In the current trial there were only 3 polyps with high-grade dysplasia, a number too small for further analysis.

Discussion

The procedure characteristics indicated that water exchange was appropriately applied (Table 2). Despite site-specific variations in published baseline water exchange ADR (Table 1), the aggregated data confirmed the efficacy of the novel method: i.e. the ADR in all study groups of water exchange, with or without selected cap (Table 3), surpassed the benchmark (35%) [16]. A more comprehensive metric that may be less prone to variability and reflects more thorough mucosal inspections is APC [18], which measures the average number of adenomas found per procedure. The proposed benchmarks are 0.46-0.50 for men and 0.13-0.20 for women [18]. Our APC data surpassed the benchmark (Table 3): No cap 0.991±1.522, CAP 1.119±1.637, Daisycuff^™ 1.094±1.473 and Endocuff Vision^® 1.313±1.618.

Logistic regression analysis revealed several factors correlated with high APC (Table 4), confirmed their importance [19-21], and enhanced the credibility of the novel cap findings. There were variations in site-specific pretrial ADR (Table 1) and current ADR outcomes (Supplementary Table 2A) and a lack of site-based cap effects on ADR (Supplementary Table 2B). There was a lack of significant influence of colonoscopy indications on ADR, except for the FIT+ indication (Supplementary Tables 3A-C). These variations were evened out by the inclusion of multiple investigators, whose outcome data were aggregated to yield the above benchmark ADR and APC (Table 3).

The effect of sedation type on ADR is variable. No significant difference was shown in ADR between conscious vs. deep sedation (50% vs. 54%; P=0.394) [22], and between moderate vs. deep sedation (35.9% vs. 37.3%; P=0.82) [23]. A study that included 196 endoscopists and 52,506 patients revealed that the anesthesia state did not enhance the ADR [24]. Some findings even suggested that the ADR may be higher for endoscopy without anesthesia [25]. While the use of propofol-based anesthesia was associated with better patient satisfaction and pain levels, ADR was not enhanced [26]. In the current study, sedation type had no significant impact on the effect of caps on ADR (Supplementary Table 4), except that the analysis of data from 5 sites found that CAP showed a significantly higher ADR than No cap in unsedated subjects (P=0.008; 95%CI 0.25-0.463).

Large colon polyps (>10 mm in size) account for about 5% of all colon polyps. Their presence indicates a higher risk of cancer development. The prevalence of large polyps was reported to be in the range of 6-7% [27]. The earliest water exchange report described 16.7% of patients with large (>10 mm) polyps [8]. The current study found the following large polyp detection rates (Supplementary Table 5): No cap (8.6%), CAP (8.6%), Daisycuff^™ (6.9%), Endocuff Vision^® (14.4%), all higher than the values reported by Leiberman et al [27]. The current data show that the detection rates of diminutive and small polyps were consistent with reported data [27-30]. The detection rate of diminutive and small polyps was not affected by the caps (Supplementary Tables 6 and 7).

The proportion of flat polyps varies by study. One report indicated that 24.2% of polyps could be flat [29]. While not the most common shape, flat polyps are important, because they are harder to find and remove completely during a colonoscopy and may carry a higher risk of containing cancer or high-grade dysplasia. The detection rates of flat polyps (Supplementary Tables 8 and 9) in the present study were comparable to published data, and were not affected by the caps.

Serrated polyps have variable definitions, which continue to evolve. The sessile serrated polyp detection rate has been reported at 3.3-5.1% [31]. One review reported a sessile serrated polyp detection rate of 2.5% (1.5-3.8%) [32] and another 3.3% (2.2-4.8%) [33]. A serrated polyp detection rate of 3% was linked to the development of interval cancers and considered as minimum cutoff point of competency, but striving for higher rates is critical for reducing interval colorectal cancer risk and improving colonoscopy quality [34]. The detection rates of serrated polyps in the current study were of a comparable order of magnitude (Supplementary Table 10): No cap (3.02%), CAP (4.4%), Daisycuff^™ (5.57%), Endocuff Vision^® (2.05%). The study methods provided benchmark performance but did not appear to modify the detection rates. The proportion of colon polyps with dysplasia varies, but studies show that roughly 20-30% have high-grade dysplasia [30]. We found only 3 mentions of “high-grade dysplasia” in our data, insufficient for further analysis.

This study was carried out by endoscopists at sites with a record of improving ADR using water exchange [8-11]. There were nearly equivalent mean volumes of water infused and suctioned during insertion (Table 2). Compared with the first description of water exchange, the mean procedure time improved (decreased) from 56 min [9] to ~25 min (Table 2). All withdrawals exceeded 9 min [35] and were close to 13 min [19]. The CAP protruded 2 mm beyond the tip of the colonoscope, restricted peripheral vision and possibly prolonged the withdrawal time, although the difference did not reach significance (Table 2).

In a recent meta-analysis, pooled ADRs in studies with and without a second forward view were 26% and 18%, respectively (significant difference); pooled advanced ADRs were 3.7% and 2.5% (no significant difference) [14]. In the current study a second look was not used to avoid confounding the ADR findings.

The ADR benchmark served as a key quality indicator [3], and as an aspirational goal for low detectors. Variability in ADR in control groups of randomized controlled trials necessitated standardization of clinical, methodological and technical parameters for comparisons [36]. We analyzed 6 trials (2010-2017) that compared gas (air or CO₂) insufflation (n=2699) and water exchange (n=2708). The sites were community hospitals in Mainland China (n=6), Taiwan (n=2), Europe (n=5), and Veterans Affairs centers in the United States (n=3). Water exchange showed significantly better ADR compared with air insufflation (27.4% vs. 20.9%, P=0.001) [37]. The demographics of the current study (Supplementary Table 1) mimicked those of the pooled data [37]. The No cap ADR of 45.6% (7 sites) and 44.9% (5 sites) exceeded that (27.4%) in the pooled data [37]. Only 1 site (22.7%) (Supplementary Table 2) was below the pooled data. While endoscopist experience, patient mix or local variations might have accounted for the difference, the current result of total procedure time of about 25 min (Table 2), decreased from 56 min when water exchange was first described (8) indicated that the performance of the water exchange method had improved and become more efficient.

When the entire cohort in the CAP and the No cap groups were evaluated (data from 7 sites), even though CAP did not significantly increase ADR, the CAP ADR (49.3%) exceeded the new benchmark of 35% [16]. It was plausible that the insertion cleaning of water exchange optimized the ADR results of CAP, resolving the conflicting data [5,6] due to the interference of residual fecal matter lodged in the CAP [6].

The data of the 5 sites that randomized No cap, CAP and Daisycuff^™ showed that the effect of Daisycuff^™ (52.8% vs. 44.9%) was nearly significantly P=0.05 (-16%, 0.3%). A study using larger samples is needed to overcome a type II error. A plausible underlying mechanism for high ADR could be as follows. With 2 Daisycuff^™ holding onto the pleated colon at 2 separate points (20 cm apart) along the shaft of the colonoscope, there was a smaller “bowstring effect” [38]. On withdrawal, there was less “fly-off” of the fold compressed by the Daisycuff^™ at the tip. A more controlled and stable fold exposure was provided for inspection. The more stable inspection resulted in a higher ADR and APC (Table 3C).

When the data from 2 sites were analyzed, the lack of a significant effect of Endocuff Vision^® P=0.057 (-22.9%, 0.9%) could be a type II error. Reports from the United Kingdom and Canada showed that Endocuff Vision^® increased ADR (without water exchange) from 36.2% to 40.9% (P=0.02) [39], and from 46.7% to 54.6% (P=0.001) [40], respectively. A third international report, however, showed that Endocuff Vision^® did not significantly improve ADR (41.1% vs. 35.5%; P=0.125) [41]. The ADR of 58.0% and APC of 1.313 suggest that water exchange optimized the performance of Endocuff Vision^® in adenoma detection (Table 3E).

Historical data showed that water exchange increased adenoma [11] and flat polyp [42] detection rates, as well as decreasing the adenoma miss rate [43] in the right colon. Now, CAP (Table 3B) and Daisycuff^™ (Table 3D) significantly increased the right colon ADR. The higher proportion of interval than non-interval cancers residing in the right colon [44] suggests that water exchange combined with CAP or Daisycuff^™ may have the potential to preferentially attenuate interval cancers in the right colon. The current novel findings deserve further study for confirmation. The enhancement of APC in the left colon by Daisycuff^™ (Table 3D) adds to its potential utility.

The strengths of the current study included multi-national, multi-site and multi-investigator design, effective randomization, intention-to-treat analysis, a low (<6%) proportion of poor bowel preparation, and a high (>95%) cecal intubation rate. The novel results in the exploratory assessment of Daisycuff^™ and Endocuff Vision^® in water exchange colonoscopy open new avenues for investigation into approaches to enhance the performance of water exchange colonoscopy.

The limitations included no definitive outcome data to support the proposed hypothesis that selected cap(s) improved the overall ADR performance of water exchange. In addition, some data were missing from the aggregated data file, while the uneven number of subjects randomized to each arm, as a result of the limited access to Daisycuff^™ and the substitution of Endocuff Vision^® at 2 study sites, precluded an explanation of the difference in outcome between sites. Only unblinded investigators with expertise in water exchange participated in the study. Fold exposure and interval cancers were not evaluated. The average insertion time was long (~12 min) and could remain as a disincentive in routine clinical practice. Despite the limitations, interesting pilot data (Daisycuff^™ and Endocuff Vision^®) were obtained.

In summary, this study confirmed the efficacy of water exchange, in that the ADR and APC in all study groups surpassed the benchmarks. CAP and Daisycuff^™ significantly increased ADR in the right colon. Daisycuff^™ produced a nearly significant increase in overall ADR. In conclusion, the significantly higher ADR in the right colon when CAP and Daisycuff^™ were compared with No cap suggested that they hold the promise of reducing interval cancers in the right colon. The nearly significantly higher ADR comparing Daisycuff^™ and No cap and other comparisons with No cap deserve to be reevaluated with larger samples. Confirmation of the significant positive findings of selected cap(s) may support their role in performance improvement. The combination of water exchange with these selected caps is new, and the findings should be confirmed in further studies before their use is recommended for performance improvement to increase ADR and APC.