The efficacy of colonoscopy for adenoma detection

Evidence for the efficacy of colonoscopy in detecting colorectal neoplasia derives from clinical trials of screening colonoscopy. The polyp prevalence rates calculated from the trials in the United States conducted in asymptomatic, average-risk persons aged 50 years and older have yielded fairly consistent results, with mean adenoma prevalence ranging from 25% to 37%. These prevalence rates have formed the basis of the current quality indicator thresholds for adenoma detection.

However, in this article it is argued that the true efficacy of colonoscopy for adenoma detection has not been defined, while colonoscopy has been applied in routine clinical practice with suboptimal effectiveness for adenoma detection and CRC prevention. This argument is supported by increasing observational evidence that colonoscopy does not confer the high levels of protection against incident CRC that it was once thought it could. Whereas early studies showed that colonoscopy and polypectomy prevent 76% to 90% of CRC, data now suggest that the level of protection from colonoscopy may be only 50% or less, with protection perhaps confined to the left colon. Failures of colonoscopy to consistently detect adenomas or other precursors of CRC are therefore threatening the effectiveness of colonoscopy for the prevention of CRC.

Why is the true efficacy of colonoscopy undefined? First, studies of the efficacy of colonoscopy in detecting adenomas are limited by the lack of a criterion/gold standard against which the performance characteristics of the test can be evaluated. Second, most of the so-called efficacy studies were conducted in routine clinical academic or community practice, in which the testing conditions and participants were not necessarily optimized or standardized. Therefore, these efficacy studies were subject to the same factors that influence the effectiveness of colonoscopy in clinical practice.

Factors limiting the effectiveness of colonoscopy

Limitations to the effectiveness of colonoscopy may be related to the patient, endoscopist, system, or technical factors ( Box 1 ). Patient factors include the quality of bowel preparation and tumor biology. Bowel preparation has been shown several times to contribute to polyp detection, and the use of split-dose bowel preparations is now established as a critical element of quality colonoscopy. Bowel preparation quality appears to be affected by socioeconomic status, and lower levels of bowel preparation might contribute to the failure of colonoscopists serving indigent patients to achieve target thresholds. Tumor biology is also relevant, given the increasing recognition of flat lesions and the serrated pathway of carcinogenesis.

However, endoscopist-related operator dependency in colonoscopy performance is the major factor that threatens the effectiveness of colonoscopy in routine practice. Colonoscopy may be subject to operator dependency more than any other CRC prevention test. Many gastroenterologists have been documented to have adenoma detection rates less than the current targets. The magnitude of within-group variation is substantial (4- to 10-fold) but is comparable when examined in academic, mixed community-academic, and community practices. Differences in detection extend importantly to detection of large adenomas. Given that colonoscopy plays a critical role in CRC prevention when it is used either for screening or for diagnosis with other positive screening tests and given the widespread acceptance of the importance of CRC prevention, the overall injury caused by suboptimal effectiveness of colonoscopy is substantial.

Between-specialty variation has also been documented in colonoscopy performance. Primary care physicians performing colonoscopy have been shown several times to be more likely to miss CRC when compared with gastroenterologists, and in Canada, performance of colonoscopy in the office setting is associated with an increased risk of missed cancer diagnosis. These differences may reflect the differences in training in colonoscopy between primary care physicians and gastroenterologists. However, the purpose of developing quality targets for colonoscopy is to reduce variation in performance, that is, to bring all performance colonoscopy up to minimum accepted thresholds. Thus, although thresholds are derived from studies performed by gastroenterologists, quality targets derived from those studies are available for use and should be used by general surgeons and primary care physicians performing colonoscopy.

The potential negative effect of ineffective colonoscopy is apparent when considering the current postpolypectomy surveillance recommendations. Postpolypectomy surveillance recommendations, by necessity, assume that the performance of colonoscopy is relatively uniform. In reality, the percentage of patients with one or more adenomas, with multiple adenomas, and with large adenomas differs substantially between endoscopists. If endoscopists have low adenoma detection rates, they will incorrectly advise more adenoma-bearing patients that they are normal and can be followed up at 10 years, and more patients with 3 or more adenomas that they can be followed up at 5 years. Thus, a low-level adenoma detector following the current postpolypectomy surveillance recommendations would be expected to put a significant percentage of their patients at increased risk of an interval cancer. On the other hand, a high-level adenoma detector’s patients are doubly protected. First, their patients’ colons have been better cleared of adenomas at the baseline colonoscopy. Second, they are reassessing substantially more patients at 5 and 3 years’ follow-up because of the presence of any adenoma, or of multiple (3 or more) adenomas and large adenomas respectively. Many colonoscopists do not follow postpolypectomy surveillance recommendations, perhaps because of lack of knowledge regarding the recommendations or lack of confidence in the recommendations. Undoubtedly, many of these same colonoscopists have not measured their personal adenoma detection rates. Concern from endoscopists who have not measured their adenoma detection rate that current screening and surveillance recommendation intervals are too long is illogical and arguably irresponsible, because only a knowledge of the adenoma detection rate allows an estimate of whether an endoscopist is likely to be placing patients at risk for missed lesions.

Technical factors include the availability of quality endoscopic equipment, and system factors include those financial or organizational factors that may preclude or work against best practice. High-resolution white light colonoscopy has become widely available since the publication of adenoma detection targets, probably facilitating adenoma detection, and therefore technology is unlikely to contribute to lower detection rates in clinical practice. Further technical innovations in colonoscopic imaging should make reaching adenoma detection targets even easier. In this regard, high-definition colonoscopy shows promise, although other techniques such as pancolonic indigo carmine chromoendoscopy, narrow band imaging, and the Fuji Intelligent Chromo Endoscopy system (Fujinon Inc, Wayne, NJ, USA) have generally yielded negative results regarding improvement in adenoma detection.

Factors limiting the effectiveness of colonoscopy

Limitations to the effectiveness of colonoscopy may be related to the patient, endoscopist, system, or technical factors ( Box 1 ). Patient factors include the quality of bowel preparation and tumor biology. Bowel preparation has been shown several times to contribute to polyp detection, and the use of split-dose bowel preparations is now established as a critical element of quality colonoscopy. Bowel preparation quality appears to be affected by socioeconomic status, and lower levels of bowel preparation might contribute to the failure of colonoscopists serving indigent patients to achieve target thresholds. Tumor biology is also relevant, given the increasing recognition of flat lesions and the serrated pathway of carcinogenesis.

However, endoscopist-related operator dependency in colonoscopy performance is the major factor that threatens the effectiveness of colonoscopy in routine practice. Colonoscopy may be subject to operator dependency more than any other CRC prevention test. Many gastroenterologists have been documented to have adenoma detection rates less than the current targets. The magnitude of within-group variation is substantial (4- to 10-fold) but is comparable when examined in academic, mixed community-academic, and community practices. Differences in detection extend importantly to detection of large adenomas. Given that colonoscopy plays a critical role in CRC prevention when it is used either for screening or for diagnosis with other positive screening tests and given the widespread acceptance of the importance of CRC prevention, the overall injury caused by suboptimal effectiveness of colonoscopy is substantial.

Between-specialty variation has also been documented in colonoscopy performance. Primary care physicians performing colonoscopy have been shown several times to be more likely to miss CRC when compared with gastroenterologists, and in Canada, performance of colonoscopy in the office setting is associated with an increased risk of missed cancer diagnosis. These differences may reflect the differences in training in colonoscopy between primary care physicians and gastroenterologists. However, the purpose of developing quality targets for colonoscopy is to reduce variation in performance, that is, to bring all performance colonoscopy up to minimum accepted thresholds. Thus, although thresholds are derived from studies performed by gastroenterologists, quality targets derived from those studies are available for use and should be used by general surgeons and primary care physicians performing colonoscopy.

The potential negative effect of ineffective colonoscopy is apparent when considering the current postpolypectomy surveillance recommendations. Postpolypectomy surveillance recommendations, by necessity, assume that the performance of colonoscopy is relatively uniform. In reality, the percentage of patients with one or more adenomas, with multiple adenomas, and with large adenomas differs substantially between endoscopists. If endoscopists have low adenoma detection rates, they will incorrectly advise more adenoma-bearing patients that they are normal and can be followed up at 10 years, and more patients with 3 or more adenomas that they can be followed up at 5 years. Thus, a low-level adenoma detector following the current postpolypectomy surveillance recommendations would be expected to put a significant percentage of their patients at increased risk of an interval cancer. On the other hand, a high-level adenoma detector’s patients are doubly protected. First, their patients’ colons have been better cleared of adenomas at the baseline colonoscopy. Second, they are reassessing substantially more patients at 5 and 3 years’ follow-up because of the presence of any adenoma, or of multiple (3 or more) adenomas and large adenomas respectively. Many colonoscopists do not follow postpolypectomy surveillance recommendations, perhaps because of lack of knowledge regarding the recommendations or lack of confidence in the recommendations. Undoubtedly, many of these same colonoscopists have not measured their personal adenoma detection rates. Concern from endoscopists who have not measured their adenoma detection rate that current screening and surveillance recommendation intervals are too long is illogical and arguably irresponsible, because only a knowledge of the adenoma detection rate allows an estimate of whether an endoscopist is likely to be placing patients at risk for missed lesions.

Technical factors include the availability of quality endoscopic equipment, and system factors include those financial or organizational factors that may preclude or work against best practice. High-resolution white light colonoscopy has become widely available since the publication of adenoma detection targets, probably facilitating adenoma detection, and therefore technology is unlikely to contribute to lower detection rates in clinical practice. Further technical innovations in colonoscopic imaging should make reaching adenoma detection targets even easier. In this regard, high-definition colonoscopy shows promise, although other techniques such as pancolonic indigo carmine chromoendoscopy, narrow band imaging, and the Fuji Intelligent Chromo Endoscopy system (Fujinon Inc, Wayne, NJ, USA) have generally yielded negative results regarding improvement in adenoma detection.

Defining the true efficacy of colonoscopy

Recommended thresholds for adenoma detection in asymptomatic persons aged 50 years and older are at present set at 25% for men and at 15% for women. These targets were set by consideration of the mean prevalence rates of adenomas in screening colonoscopy studies. Thus, the mean adenoma prevalence rates in all screening colonoscopy studies in the United States have exceeded the recommended thresholds of 25% in men and 15% in women. However, the same factors that limit the effectiveness of colonoscopy have also likely influenced the conduction of the original studies that have established the current definitions of the efficacy of colonoscopy.

In these studies, patient characteristics were not standardized. For example, the prevalence of adenomas is influenced not only by the age and gender of the study population but also by additional factors such as the prevalence of smoking, obesity, diabetes, and a family history of CRC. As a result, it is important to be cautious in generalizing the adenoma prevalence findings from these trials, because there are insufficient data present to determine whether the study populations were typical of patients assessed in routine screening colonoscopy practice. However, because volunteers for studies are generally healthier than the general population, it is unlikely that differences between populations in clinical trials and clinical practice would result in practitioners being less likely to be able to achieve targeted neoplasia detection rates. Other patient factors, such as the type and quality of bowel preparation, which have a clear influence on polyp detection, were not standardized in the efficacy studies. Split-dose preparations were not widely used at that time and tended to equalize differences in preparation quality between populations.

Regarding endoscopist factors, the efficacy studies did not report variation in detection between the participating colonoscopists. However, there is every reason to think that the same level of variation between colonoscopists was present in these studies as has been found in other studies describing variable adenoma detection. Technical factors are also likely to have influenced the measured efficacy in screening studies. Efficacy studies of screening colonoscopy are generally characterized by the use of white light endoscopy and standard-resolution colonoscopes, with high cecal intubation rates of 97% and more.

Recent studies in which test conditions have been optimized have redefined the efficacy of colonoscopy for adenoma detection. These studies have shown that known high-level adenoma detectors using high-definition colonoscopes with high-quality bowel preparation find adenomas in 50% or more of the screening population. In most cases, the gain in adenoma detection is in diminutive and frequently in flat adenomas. As a result, the costs and risks associated with removal of many diminutive adenomas assume increasing importance with increasing adenoma detection, and methods for real-time histologic assessment to allow small and diminutive adenomas to be “resected and discarded” are needed.

Ultimately, these high rates of true adenoma prevalence could be used to increase adenoma detection targets. However, increasing targets raises the question of whether thresholds should be set at the arithmetical mean of all the studies, or perhaps the mean determined in a weighted fashion by the numbers of subjects in the studies, or at a different level, such as one-third (or 10%) of the mean adenoma prevalence rates in the studies. In this regard, the recommended adenoma detection thresholds are arbitrary. However, the current thresholds establish a cutoff level such that individuals with detection levels below the thresholds are definitely in the lowest half of adenoma detectors and are detecting less than half of those with adenomas. Given that many endoscopists have not yet measured their adenoma detection rates at all, it seems unnecessary to raise detection targets at this time, although this could be considered in the future.

In addition to variation in the documented prevalence rates of adenomas in screening colonoscopy, it is important to acknowledge the limitations of the adenoma detection rate as a definition of efficacy. The rate of adenoma detection in screening colonoscopy is only an intermediate outcome measure compared with the most definitive end-outcome measure, which is the incidence of interval cancer after clearing colonoscopy. Interval cancer incidence has been shown statistically to cluster within individual colonoscopists, which would be expected given the current information about variable detection between endoscopists. However, adenoma detection is probably the best marker of the quality of mucosal inspection available.

Large adenoma detection is logically a superior outcome measure to overall adenoma detection, but is problematic to measure. The acceptable target thresholds for large adenoma detection are less clearly based on screening colonoscopy studies. Second, measurement of the large-adenoma detection rate is more difficult because many more colonoscopies must be reviewed to determine the large-adenoma detection rate with narrow confidence intervals, simply because they are much less prevalent than adenomas in general. Third, the use of a large-adenoma detection rate introduces an error caused by endoscopic polyp size measurement. In addition, measurement bias could be introduced by endoscopists who simply designate polyps less than 1 cm in size as 1 cm or larger to reach the detection rate. Overall adenoma detection is not subject to this bias.

The best proposed alternative to the adenoma detection rate is the detection of number of adenomas per colonoscopy, because it might best reflect the quality of mucosal examination over the entire length of the colon and have increased discriminatory power for good versus poor detectors. However, the best thresholds for number of adenomas detected per colonoscopy are not yet known. Thus, overall adenoma detection defined as a prevalence rate, that is, the percentage of patients aged 50 years and older with one or more adenomas, is likely to remain the single most important and feasible measure of colonoscopy efficacy. Overall adenoma detection has the advantage of being easy to measure and having been well studied in efficacy trials, and at least 2 studies show that it correlates with large adenoma detection.