All endoscopes are equal, but some are more equal than others

Two of the key aims of diagnostic colonoscopy are to identify colorectal cancer (CRC) and to prevent CRC by detecting and resecting premalignant polyps. However, it is an imperfect test and patients present with post-colonoscopy CRC due to missed lesions [1]. To minimize this risk, the entire colorectal mucosa needs to be carefully inspected. The detection process is multifaceted.

“… it is not what you use but rather the way that you use it ... so pick your endoscopist before you pick your technology.”

First, the mucosa needs to be prepared and unmasked. This process starts before the procedure, with high quality bowel preparation: studies consistently show that this is best delivered as a split dose, but there is more to it than that, including adequate fluid intake, discontinuation of certain medications, and dietary modification. Optimal peri-procedural unmasking of the mucosa involves additional irrigation as required, position shifts to open up the lumen, and adequate gas insufflation. Buscopan may improve visualization by reducing haustral tone, although the evidence for this is mixed. Endoscopic accessories such as cuffs or rings can also be used to flatten the folds.

Second, a careful, active extubation technique is required. There is a well-established correlation between inspection time and pathology detection rate. Although a minimum inspection time of 6 minutes is recognized, the optimal inspection time is probably at least 10 minutes [2]. However, how the endoscopist spends that time is equally important, and involves segmental re-intubation when endoscope slippage occurs, to ensure that all the mucosa is visualized, especially in potential blind spots such as the proximal aspects of flexures and folds. Studies show that some endoscopists are more active at watching the screen than others, scanning the peripheries of the screen rather than just looking at the middle, or worse still being distracted and spending more time looking away from the screen than at it.

Third, the “system acuity” should be as high as possible. In theory, higher definition endoscopes should improve one’s ability to detect pathology that is present on the now-unmasked mucosa. Mucosal illumination needs to be sufficient to allow both near- and far-field illumination without glare; this was an issue with early versions of narrow-band imaging, where the light brightness was insufficient to allow far-field illumination. Next, the screen should also be high definition, of sufficient size, and appropriately positioned in the room to exploit the high definition images. The endoscopist’s visual acuity must also not be compromised. Of course, if the mucosa has not been adequately unmasked in the first place, then high definition equipment is fairly pointless.

Fourth, the endoscopist needs to have adequate knowledge of what is being looked for and experience of pattern recognition, picking up subtle clues that pathology may be present. A good illustration of this is in the evolution of the sessile serrated lesion story: these lesions are subtle and it is only with greater understanding of their appearance (pallor, cloud-like surface, indistinct borders often capped with bile-stained mucus) and knowledge that such lesions are clinically significant, that we have begun to detect them with increasing frequency – to paraphrase D.H. Lawrence, “the eye doesn’t see what the mind doesn’t know.”

A chain is only as strong as its weakest link and minimizing pathology miss rates requires all of the above aspects to be optimized.

In this issue of Endoscopy, Roelandt et al. present a randomized controlled trial looking at one aspect of the detection process: the impact on adenoma detection rate (ADR) of standard definition or high definition endoscopes, both with white-light colonoscopy and with virtual chromoendoscopy [3]. It is a large trial (although they failed to reach their recruitment target), optimistically powered to detect an absolute ADR difference of 10 %. The study failed to show a significant difference. This may be due to type II error (as both high definition white-light colonoscopy and high definition virtual chromoendoscopy ADRs were numerically higher than their counterparts); alternatively, there may genuinely be no difference. I am inclined to believe the former: it is always difficult to detect differences in incrementally improving technologies – nobody would doubt that we can see more now than in the fiberoptic era, but subtle improvements in definition will, at best, result in subtle increases in detection rates, requiring even larger studies to detect the difference. It is akin to parents barely noticing that their child has grown while the great aunt will always see the change. Another explanation may be that high definition does permit the detection of more subtle pathology, but once an endoscopist has developed an understanding of the subtle clues they can translate this back to standard definition views, paying more attention to areas they may previously have ignored.

I do wonder whether the authors have used the most appropriate primary end point: presenting mean number of adenomas would be more logical than an “all or none” ADR, given they are assessing the ability of the technologies to detect pathology. Although they present “adenomas per colonoscopy rate,” their methodology indicates that this is actually “mean adenomas per positive procedure” – an important distinction. They also include sessile serrated lesions in their adenoma count: it would be preferable to analyze adenomas separately. Some of this can be inferred and shows that virtual chromoendoscopy (both high and standard definition) performs numerically worse than white-light colonoscopy (I am tempted to believe this, as I see virtual chromoendoscopy as a lesion characterization rather than a detection technology), and that standard definition (white-light colonoscopy or virtual chromoendoscopy) performs numerically better than high definition, driven by the number of diminutive lesions detected – something they do mention, offering a potential explanation that edge definition is better with high definition, leading to an over-estimation (or possibly underestimation with standard definition) of polyp size. An alternative explanation would be that some unknown confounder or bias has encroached into their study. One issue with all endoscopy trials is that it is impossible to blind the endoscopist, hence conscious or unconscious bias is always a risk. Maybe the future lies with artificial intelligence, operating as an objective arbiter of completeness of examination, the ability of technologies to detect pathology, and of endoscopist-missed pathology.

In summary, the results of this negative study throw up more questions than answers. On subgroup analysis, high definition detected more sessile serrated lesions but fewer diminutive adenomas – it is a confusing picture. What have we learned? First, these studies are challenging to run and fraught with issues – I applaud the researchers for their efforts. Second, this study would indicate that there is probably no role for the routine use of virtual chromoendoscopy as a detector technology. Third and perhaps most importantly, the study demonstrates that it is not what you use but rather the way that you use it – technology appears to account for very little of the variation in pathology detection rates that are seen – so pick your endoscopist before you pick your technology.

• References

• 1 Rutter MD, Beintaris I, Valori R. et al. World Endoscopy Organization consensus statements on post-colonoscopy and post-imaging colorectal cancer. Gastroenterology 2018; 155: 909-925
• 2 Lee TJ, Blanks RG, Rees CJ. et al. Longer mean colonoscopy withdrawal time is associated with increased adenoma detection: evidence from the Bowel Cancer Screening Programme in England. Endoscopy 2013; 45: 20-26
• 3 Roelandt P, Demedts I, Willekens H. et al. Impact of endoscopy system, high definition, and virtual chromoendoscopy in daily routine colonoscopy: a randomized trial. Endoscopy 2019; 51: 237-243