- Top of page
Background Colonoscopy has a known miss rate for polyps and adenomas. High definition (HD) colonoscopes may allow detection of subtle mucosal change, potentially aiding detection of adenomas and hyperplastic polyps.
Aim To compare detection rates between HD and standard definition (SD) colonoscopy.
Methods Prospective, cohort study with optimized withdrawal technique (withdrawal time >6 min, antispasmodic, position changes, re-examining flexures and folds). One hundred and thirty patients attending for routine colonoscopy were examined with either SD (n = 72) or HD (n = 58) colonoscopes.
Results Groups were well matched. Sixty per cent of patients had at least one adenoma detected with SD vs. 71% with HD, P = 0.20, relative risk (benefit) 1.32 (95% CI 0.85–2.04). Eighty-eight adenomas (mean ± standard deviation 1.2 ± 1.4) were detected using SD vs. 93 (1.6 ± 1.5) with HD, P = 0.12; however more nonflat, diminutive (<6 mm) adenomas were detected with HD, P = 0.03. Twenty-three proximal hyperplastic polyps (0.32 ± 0.58) were detected with SD vs. 31 (0.53 ± 0.86) with HD, P = 0.35. Overall prevalence of proximal large (>9 mm) hyperplastic polyps was 7% (0.09 ± 0.36).
Conclusions High definition did not lead to a significant increase in adenoma or hyperplastic polyp detection, but may help where comprehensive lesion detection is paramount. High detection rates appear possible with either SD or HD, when using an optimized withdrawal technique.
- Top of page
Colonoscopy provides unparalleled views of the colonic mucosal surface and has become the dominant lower gastrointestinal investigation with over 14 million examinations performed in the US per year.1 A majority of these examinations are performed for screening with the aim of detecting and removing premalignant polyps to prevent their development into colorectal carcinoma. Since the first successful total colonoscopies in the late 1960s, colonoscopic equipment has progressed rapidly from fibreoptic colonoscopes, to video colonoscopes, to high resolution [HR, increased number of pixels in charged couple device (CCD) with output to a standard resolution monitor] colonoscopes, to the current series of high definition (HD, substantially increased number of pixels in CCD with out put to a HD monitor) colonoscopes.2–4
Each new generation of instruments and their associated light sources, video processors and monitors has improved the quality of the image presented to the endoscopist. Theoretically, this should improve diagnostic accuracy, particularly for polyps, but there are no prospective, head-to-head trials of lower vs. higher definition colonoscopes. Comparisons of case series with historical controls or retrospective case–control studies either by the same endoscopist or from the same institution report conflicting results, either suggesting no difference or a large increase in adenoma detection.3, 5, 6 Certainly, a recent series using high definition (HD) colonoscopes suggested a remarkably high rate of adenoma detection in an unselected series of patients.3 It remains unclear, however, whether this reflects excellent and meticulous technique from a known high performing operator or the influence of HD allowing detection of subtle lesions. Operator performance has a very substantial effect on adenoma detection with differences of 10-fold for detection of adenomas of any size, and three- to fourfold for advanced adenomas.7, 8
Although we know there is a significant miss rate for adenomas, 22% [95% confidence interval (CI) 15–32] in back-to-back studies, it is unclear how to reduce this.9 The miss rate for hyperplastic polyps may be even higher and there has recently been considerable interest in the role of proximal hyperplastic/serrated polyps, which may lead to microsatellite unstable (MSI-high) colorectal cancer in the proximal colon via the ‘serrated pathway’.10–12 Potential endoscopic mechanisms to reduce miss rates include interventions to reveal more of the colonic surface such as cap fitted colonoscopy, wide angle colonoscopes, position changes, antispasmodics and retrograde viewing auxiliary imaging devices13–18 and interventions, which may improve surface detail and contrast and hence increase lesion detection, such as chromoendoscopy, narrow band imaging, autofluorescence imaging and higher definition white light instruments.3, 5, 19–21 However, in theory, there must be a limit at which higher definition does not allow further increases in clinically significant lesion detection, either because all lesions have been detected or because undetected lesions are not visible during white light examination.
In this study, we aimed to investigate whether the use of high definition colonoscopes improved adenoma and hyperplastic polyp detection compared to the use of standard definition (SD) colonoscopes with white light when the quality of basic withdrawal technique was maximized.
- Top of page
One hundred thirty patients completed the study, 72 in the SD and 58 in the HD groups. Demographic and clinical characteristics for the two groups are shown in Table 1. The groups were well matched with no significant differences between them. Patients were offered intravenous sedation, median dose of midazolam 1.25 mg (range 0–2.5), and pethidine 50 mg (0–75). All patients received intravenous antispasmodic [hyoscine butylbromide, median 20 mg (range 20–40), n = 127; or glucagon, 1 mg, n = 2] unless contraindicated.
Table 1. Demographic and clinical characteristics of study participants
|Variable||Standard definition, n = 72||High definition, n = 58||P-value|
|Age, median (range) [yr]||61.5 (44–80)||60 (51–79)||0.80|
|Gender, male (%)||34 (47%)||36 (62%)||0.11|
|Withdrawal time, median (range) [min:s]‡||10:15 (8:22–18:45)||10:02 (8:11–16:35)||0.70|
| Family history colorectal cancer||27 (38%)||20 (34%)|| |
| Polyp follow-up||27 (38%)||29 (50%)||0.22|
| Rectal bleeding*||5 (7%)||5 (9%)|| |
| Other†||13 (18%)||4 (7%)|| |
| Good||36 (52%)§||29 (50%)||0.86|
For the primary end point, the proportion of patients with at least one adenoma detected, there was a 11% absolute increase in the HD group that was not statistically significant, nor was the difference in the total numbers of adenomas detected, Table 2; however, the mean number of adenomas was 30% greater in the HD group and the proportion of patients with at least three adenomas was almost double that of the SD group; neither result was statistically significant. The relative risk of detection of at least one adenoma with the use of HD was 1.32 (95% CI 0.85–2.04). Combining both groups, the overall proportion of patients with at least one adenoma was 84/130 (65%), with a mean adenoma detection rate of 1.4 ± 1.5 per patient. Results for polyp numbers were similar, but differences were less pronounced between groups.
Table 2. Polyp and adenoma detection rates and numbers between standard and high definition colonoscopy polyp and adenoma numbers
|Lesion||Standard definition, n = 72||High definition, n = 58||P-value|
| Total number||88||93||0.12|
| Total mean ± s.d. || 1.2 ± 1.4|| 1.6 ± 1.5||–|
| Patients with ≥1||43 (60%)||41 (71%)||0.20|
| Patients with ≥3||13 (18%)||17 (29%)||0.l5|
| Total number||155||145||0.25|
| Total mean ± s.d.|| 2.2 ± 2.0|| 2.5 ± 2.0||–|
| Patients with ≥1||57 (79%)||47 (81%)||0.83|
| Patients with ≥3||24 (33%)||28 (48%)||0.11|
There were no significant differences in size or distribution of polyps and adenomas; however, the proportion of flat polyps was significantly higher in the SD group, being roughly double that of the HD group, P = 0.04. A similar effect was seen with the proportion of flat adenomas, but this did not reach statistical significance. There was a statistical trend to larger adenomas in the standard endoscopy group, P = 0.09 (Table 3). Consistent with this, there was almost double the proportion of adenomas ≥6 mm in the SD compared to HD group, but this difference was not significant. A sub-group analysis of adenomas 0–5 mm in size showed a statistical trend towards a greater number of lesion detected in the HD group, P = 0.06 (Table 4). When this subgroup was restricted to nonflat lesions, a significantly higher number of lesions were detected with HD, P = 0.03, with an approximately 50% increase in the mean adenoma detection rate. There was a statistical trend towards more patients having at least one nonflat, 0–5 mm adenoma detected with HD, P = 0.08. Overall, 29/51 (57%) of polyps ≥6 mm were adenomas. The proportion of advanced adenomas was similar between groups. Of a total of nine advanced (size ≥10 mm, high grade dysplasia or >20% villous elements) lesions, two (22%) were <6 mm in size (Table 3).
Table 3. Polyp and adenoma size, morphology and location
|Lesion||Standard definition||High definition||P-value|
|Adenomas||n = 88||n = 93||–|
| Size, median (range) [mm]|| 3 (1–20)|| 3 (1–15)||0.09|
| Size, mean ± s.d. [mm]|| 4.0 ± 2.9|| 3.3 ± 2.1||–|
| Size ≥6 mm (%)*||18/87 (21%)||11/92 (12%)||0.15|
| Flat (%)*||12/87 (14%)|| 6/92 (7%)||0.14|
| Proximal (%)†||60 (68%)||67 (72%)||0.63|
| Advanced (%)‡|| 5 (5.7%)|| 4 (4.3%)||0.74|
|Polyps||n = 155||n = 145||–|
| Size, median (range) [mm]|| 3 (1–20)|| 3 (1–30)||0.33|
| Size, mean ± s.d. [mm]|| 4.0 ± 3.0|| 3.8 ± 3.3||–|
| Size ≥6 mm (%)*||29/150 (19%)||23/144 (16%)||0.54|
| Flat (%)*||31/150 (21%)||16/141 (11%)||0.04|
| Proximal†||99 (63%)||102 (70%)||0.27|
Table 4. Adenomas <6 mm in size
|Lesion||Standard definition, n = 72||High definition, n = 58||P-value|
|Total mean ± s.d. ||0.97 ± 1.21||1.41 ± 1.36||–|
|Patients with ≥1||39 (54%)||39 (67%) ||0.15|
|Total nonflat mean ± s.d.||0.83 ± 1.09||1.34 ± 1.33||–|
|Patients with ≥1 nonflat*||35 (49%)||38 (66%)||0.08|
Hyperplastic polyps were detected proximal to the recto-sigmoid in 40/130 (31%) of patients, but there was no significant difference in numbers detected between groups (Table 5); however, the mean number of hyperplastic polyps detected was approximately 2/3 greater in the HD group, with a similar effect seen for mean numbers of large hyperplastic polyps. Overall, 12/130 (7%) of patients had at least one large (≥10 mm) hyperplastic polyp, with an average of 0.09 ± 0.36 per patient, with no significant differences between groups. Of the 12 large proximal hyperplastic polyps, 10 were proximal to the splenic flexure. Proximal hyperplastic polyps were significantly more likely to have a flat morphology than adenomas, 19/52 (37%) vs. 18/179 (10%), respectively, P < 0.001.
Table 5. Hyperplastic polyps proximal to the recto-sigmoid
|Lesion||Standard definition, n = 72||High definition, n = 58||P-value|
|Total mean ± s.d.|| 0.32 ± 0.58|| 0.53 ± 0.86||–|
|Patients with ≥1||20 (28%)||20 (34%)||0.45|
|Size, median (range) [mm]|| 4 (1–10)|| 5 (2–30)||0.30|
|Size, mean ± s.d. [mm]|| 4.9 ± 3.0|| 6.2 ± 5.5||–|
|Flat (%)*|| 9/23 (39%)||10/29 (34%)||0.78|
|Total number ≥10 mm (%)*|| 5/22 (22%)|| 7/31 (23%)||1.0|
|Total ≥10 mm mean ± s.d.|| 0.07 ± 0.26|| 0.12 ± 0.46||–|
|Patients ≥1 ≥10 mm|| 5 (7%)|| 4 (7%)||1.0|
- Top of page
This study prospectively compared standard and HD colonoscopy and showed a very high overall adenoma detection rate (65%) with both standard and HD colonoscopy, but no significant difference in detection rates between the techniques. Although no significant difference was seen, there were potentially clinically important increases in the total number of adenomas detected and the number of patients with at least three adenomas detected. Baseline demographic and clinical characteristics were not significantly different between the two groups; however, male patients and patients with an indication of polyp follow-up, two groups known to be more likely to have adenomas, were over-represented in the HD cohort.28, 29 These high levels of adenoma detection using optimal withdrawal technique (>6 min withdrawal time, position changes, antispasmodic, re-examining folds and flexures7, 14, 16, 24, 25, 30), 60% SD, 71% HD, are consistent with high rates seen by a known high performing operator using HD white light, 67%.3 The lack of difference between groups suggests that operator technique is possibly the more important factor in attaining high adenoma detection rates, rather than endoscopic resolution. This is consistent with two studies, one which compared HD adenoma detection with recent (median time to previous colonoscopy 24 months) paired historical examinations with SD in hereditary nonpolyposis colorectal cancer (HNPCC) surveillance and saw no difference, and another which used SD scopes where the operator with better technique had a lower adenoma miss rate.5, 25 Other studies, which have examined elements of operator technique, confirm that longer withdrawal time and position changes improve adenoma and polyp detection, and use of antispasmodic increases colonic surface visualization.7, 14, 16, 24, 25, 30 A larger randomized trial with careful matching for operators will be necessary to determine if HD colonoscopes lead to a clinically important increase in adenoma detection. Based on this study, which had a power to detect a 25% difference in the number of patients with at least one adenoma and showed no significant difference, and the observed 30% increase in mean adenoma detection rate with HD, such a study should be powered to look for differences of less than 30%.
Surprisingly we found a significantly greater proportion of flat lesions detected by SD compared with HD by a factor of almost two, which is counterintuitive as one would have expected the finer detail with HD to assist in detecting subtle, flat mucosal change. A similar but nonsignificant result was found for the proportion of flat adenomas. The higher proportion of flat lesions seen with SD appears to be partly explained by the significantly higher numbers of diminutive (0–5 mm), nonflat adenomas detected in the HD group, leading to a lower proportion of flat lesions overall with HD. This is also consistent with the smaller mean size of lesions detected with HD, with a statistical trend towards smaller median adenoma size. The biological importance of such diminutive lesions in the development of colorectal cancer is unclear, but they are generally considered to be low risk except in certain contexts, such as HNPCC where even a small lesion may rapidly develop into a carcinoma and in the detection of polyposis syndromes.31–33 Chromoendoscopy has been demonstrated to unmask polyposis where the diagnosis was previously missed, and anecdotally we have found several patients attending for polyp follow-up who, when examined with a HD instrument, clearly had an attenuated colonic polyposis syndrome that was previously undiagnosed.32 HD may therefore have a role in colonoscopic screening and surveillance where comprehensive detection of even diminutive lesions is critical. More comprehensive adenoma detection may also allow more appropriate colorectal cancer risk stratification to allocate surveillance intervals. Recent data for Lieberman et al. suggest that the risk of future high grade dysplasia or carcinoma increases rapidly with the number of adenomas detected at baseline colonoscopy.34 Patients with no adenomas had a rate of high grade dysplasia or carcinoma per 1000 person-years of follow-up of 0.7 compared to patients with three of four adenomas who had a risk of 6.6. HD appeared to help with an approximate doubling of the numbers of patients with at least three adenomas detected.
Hyperplastic polyps, particularly if large and proximal, can show histological changes and genetic alterations (BRAF mutations), consistent with a role in colorectal carcinogenesis, sometimes termed sessile serrated polyps/adenomas [SSP(A)s].12, 35 Limited data are available for the frequency of proximal hyperplastic polyps. A summary of four randomized chromoendoscopic studies suggests a prevalence of 9% with standard colonoscopy and 16% using chromoendoscopy.10 The overall prevalence rate in our study was 31%, with no difference between groups, which probably reflects meticulous technique and a new recognition of the potential importance of these lesions. The overall mean number of large (≥10 mm) hyperplastic polyps, of which on average three-quarters will be clinically significant SSP(A)s, was 0.09 per patient, with 7% of patients having at least one lesion.35 This is at the upper limit of the reported range in the literature using chromoendoscopic assisted detection, mean 0.048, range 0.01–0.1.10 This suggests that large proximal hyperplastic polyps are as or more common than large adenomas, with some authors suggesting that if they are histologically found to be SSP(A)s, they should be considered risk equivalent to adenomas.10 As hyperplastic polyps are considerably harder to detect than adenomas, this may partly explain why ‘missed’ cancers occur more commonly in the proximal colon and are more often MSI-high.10, 36 Although there was no significant difference in the number of hyperplastic polyps detected, the mean number detected with HD was two-thirds greater than that with SD. Proximal hyperplastic polyps should be included as an end point in future randomized trials, ideally with appropriate pathological expertise to sub-classify lesions as SSP(A)s.
From a histopathological viewpoint, larger studies have shown that approximately 80% of adenomas detected are 0–5 mm in size, consistent with the proportions seen in this study, with high numbers of diminutive, nonflat adenomas detected.37 Optimized technique and detection may increase pressure on histopathology services to review even greater numbers of diminutive, low risk lesions.33 One potential solution to this problem is ‘optical biopsy’ or ‘endohistology’. Three methods are available, magnification chromoendoscopy, narrow band imaging or confocal endomicroscopy, although the latter system, when integrated into a colonoscope, does not have HD optics.38–40 It seems likely that if we choose to proceed with technologies to detect smaller and smaller lesions, endoscopists may have to shoulder some of the responsibilities for diminutive lesion characterization.41
This study does have some limitations. As the study was nonrandomized (allocation on colonoscope availability), selection bias is possible, and therefore it is difficult to draw definitive conclusions. Nevertheless, the groups were relatively well-matched for baseline characteristics and the data were collected prospectively according to a standardized protocol. Also, a single endoscopist performed all procedures, which may reduce generalizability, but should ensure that operator technique was consistent throughout. As St Mark’s Hospital is a tertiary referral centre, there is probably an over-representation of patients with an indication of polyp follow-up or family history of colorectal cancer in the cohort. Furthermore, because of the double intubation methodology of the associated position change study (see Methods), it is possible that additional polyps were detected during an additional withdrawal phase where position change was not used first, which occurred in roughly half the number of cases. This could artificially increase the overall reported adenoma detection rate as seen in back-to-back miss rate studies.9 Potentially repeated examination in more than one position for each segment would further maximize lesion detection, but time constraints would probably make this approach unacceptable to busy clinicians. Although the trial size was moderate and the study could detect an absolute 25% difference in patients with at least one adenoma, smaller differences could be clinically important, particularly in the context of colonoscopic population based Bowel Cancer Screening Programs. Finally, multiple comparisons were made with the same dataset and therefore apart from the primary end point, other significant values should be considered exploratory or hypothesis-generating and viewed with care.
The effect size excluded for HD in this study was relatively small and the known effects of operator performance are large. Therefore, groups working within resource constraints with operators of variable performance may find validated interventions to improve operator performance, such as continuous quality improvement programs or training (such as the St Mark’s Accelerated Colonoscopy Training Week), more cost effective than the introduction of new technology.22, 42
Finally, it may be useful to consider which of the current techniques is likely to be the best for wide field lesion detection, chromoendoscopy, HD white light endoscopy or narrow band imaging, in routine cases. Randomized data examining HD white light endoscopy vs. HD narrow band imaging suggest no difference in detection rates for a known high performing operator.3 Chromoendoscopy, although offering benefit for diminutive lesion detection, is probably too time-consuming for routine clinical use.19, 20, 42 Therefore, continued use of white light with optimized technique is probably currently justified in routine cases. Further, larger studies will be needed to demonstrate a clinically and statistically significant advantage of HD colonoscopy to justify changing to this technology. Optimization of operator technique may be of greater relevance and is a prerequisite before taking on advanced imaging techniques.25
In summary, in this prospective cohort study, HD colonoscopy did not significantly increase adenoma or hyperplastic polyp detection compared to SD and a relatively modest effect size was excluded. High detection rates of adenomas and large proximal hyperplastic polyps appear possible with standard or HD colonoscopy with optimal operator technique. HD colonoscopy may have a role in cases where comprehensive detection of even diminutive lesions is paramount.