PDF icon Download PDF
Endoscopy 2013; 45(07): 513-515
DOI: 10.1055/s-0033-1344248
Editorial
© Georg Thieme Verlag KG Stuttgart · New York

Can high resolution microendoscopy improve the resect and discard strategy?

C. Hassan
1   Gastroenterology Unit, Nuovo Regina Margherita Hospital, Rome, Italy
,
J. E. East
2   Translational Gastroenterology Unit, John Radcliffe Hospital, Oxford, UK
› Author Affiliations
Further Information

Publication History

Publication Date:
25 June 2013 (online)

Also available ateRef
   Permissions and Reprints
Preview

Polypectomy is a major determinant of the cost of colonoscopy screening [1]. This appears to be related to the high prevalence of polyps in the general population. It has recently been estimated that up to 42 % of screenees will present with at least one polyp at screening colonoscopy [2]. This rate is likely to increase with the widespread implementation of high definition endoscopy and of quality assurance programs for screening colonoscopy [3]. Polypectomy costs are partly related to the cost of pathological examination. With regard to diminutive polyps (≤ 5 mm in diameter), which represent more than 60 % of all polyps detected by colonoscopy during average-risk screening [2], the additional cost of post-polypectomy pathological examination is mainly justified by the need to differentiate between precancerous adenomatous and hyperplastic polyps. To reduce such costs, a policy of discarding the polyp and avoiding the expense of pathological examination has been proposed; this has been termed the resect and discard policy [4] [5] [6]. Its main assumption is that if endoscopy were to allow in vivo prediction of the histology of the polyp, the additional value of pathological examination could become marginal, such examination then being inefficient and/or cost-ineffective [1] [7] [8].

The application of the resect and discard strategy in Western countries is mainly based on the use of narrowed spectrum endoscopy (narrow band imaging [NBI], Fuji intelligent color enhancement [FICE], i-SCAN), also described as electronic chromoendoscopy (ECE) [4] [5] [6] [9]. The main advantages of electronic chromoendoscopy are its simple and immediate activation, and wide availability on latest-generation endoscopy systems. ECE primarily exploits the neo-angiogenesis of neoplastic lesions in order to differentiate between adenomatous and hyperplastic lesions, by means of light filtering that is applied pre- or post-processing [10] [11] [12]. Although tertiary centers have demonstrated high accuracy of ECE for in vivo prediction of polyp histology [4] [5] [6], community-based studies have reported suboptimal results for sensitivity and specificity [13] [14] [15] Suboptimal specificity would result in unnecessary surveillance examinations for those with false-positive findings (i. e., with only hyperplastic lesions), undermining the potential economic advantage of the resect and discard strategy. Conversely, low sensitivity results in surveillance not being offered to some in whom it may actually be appropriate, risking interval cancer.

In this issue of Endoscopy, Chang and colleagues report their validation of a new advanced imaging technique for characterizing polyp histology, namely high resolution microendoscopy (HRME) [13]. In contrast to ECE, HRME provides imaging of subcellular features of colorectal glands, allowing in vivo histological assessment of the lesion [14]. Thus, discrimination between neoplastic and non-neoplastic lesions is mainly based on nuclear appearance and cell size/density rather than on their vascularization. Because of a lack of data in this field, the authors exploited pathological know-how to develop a simple HRME-based classification of the different polyp histologies, including hyperplastic, different degrees of adenoma, and cancerous [13]. They then validated this classification by showing an adequate degree of interobserver concordance among endoscopists who were expert and nonexpert regarding HRME. Finally, they provided a preliminary assessment of the accuracy of HRME in discriminating between hyperplastic and adenomatous polyps, showing a very high specificity (94 %), albeit accompanied by a suboptimal sensitivity (70 %) [13]. This needs to be viewed in the context of the “lesions” assessed: a third of these were normal mucosa, and a further third were hyperplastic, in contrast to routine polyp detection where the majority of polyps are adenomatous, especially if nondiminutive.

The main strengths of the study by Chang et al. are firstly the very high degree of concordance among endoscopists of different experience in interpreting HRME images [13]. This indicates that HRME results may be reproducible. Secondly, the authors showed that a short training period is sufficient to achieve adequate competence in polyp discrimination using HRME, similar to that reported for ECE [15]. Indeed, the post-training accuracy values reached by the endoscopists who were not experts in HRME were not inferior to those shown by the HRME experts [13]. This would in turn support the generalizability of HRME within the endoscopist community. Thirdly, HRME seems to be a highly specific technique [13]. Although the actual reasons for such a high specificity are unknown, it could be hypothesized that the subcellular characteristics of hyperplastic polyps are so distinctive that they may be readily recognized by endoscopists after a dedicated training. It will be interesting to assess the ability of HMRE to discriminate between hyperplastic polyps and sessile serrated adenomas/polyps, a task that remains problematic with ECE [16]. Fourthly, although not described in the present study, Chang et al. have also included in their new classification the possibility for differentiating between nonadvanced adenomas and advanced adenomas with a villous component [13]. Finally, Chang et al. limited the analysis to high quality images, the level of confidence having been closely associated with a higher accuracy in previous studies using ECE [5] [6] [13] [17].

Has HRME a role in a resect and discard strategy? As proposed by Chang et al., HRME has the potential to complement prediction by ECE for the following reasons [13]. First, HRME and ECE exploit different characteristics of the lesions, namely subcellular patterns and vascularization, providing complementary rather than duplicate information [5] [13]. Secondly, HRME could be limited to ECE-positive cases, thereby making optimal use of the high specificity of the technique, while minimizing the actual number of cases in which it needs to be applied. Of note, a large study recently showed a very low specificity of 40 %, when an ECE-based resect and discard strategy was implemented by community-based endoscopists [18]. Thus, exploiting the high specificity of HRME could prevent substantial overuse of endoscopic surveillance by correctly reclassifying some screenees as true negatives (i. e., without adenomatous lesions). Thirdly, HRME shares with ECE the advantages of being relatively cheap and easily available, so that the addition of HRME to ECE appears technically and economically feasible. Finally, a main limitation of ECE evaluation is an inability to discriminate between advanced and nonadvanced adenomas; this means that a few screenees with diminutive but nevertheless advanced adenomas would mistakenly not be offered appropriate surveillance. If the validity of Chang et al.’s classification were to be confirmed in future studies, HRME could help to avoid an overlong surveillance interval in those with tubulovillous diminutive lesions, simply characterized as “adenomatous” at ECE, by correctly reclassifying such lesions as “advanced”.

Does this mean that the time is right for implementation of HRME in clinical practice? Chang et al. provided only a preliminary analysis [13], and several issues still need to be addressed. First, as a topical contrast agent to show nuclear features, the authors used proflavine, a drug currently in clinical use as a topical antiseptic, under an FDA application as an investigational new drug; however there have been concerns that proflavine might be carcinogenic [19]. Although low concentrations and volumes of proflavine were used, this issue must be definitively resolved before widespread human use. Secondly, although the stored images were obtained in vivo from real patients, all the concordance and accuracy values were derived from the artificial setting of ex vivo analysis [13]. In previous experiences with ECE-based prediction the optimal performances achieved ex vivo have not been confirmed in vivo [18]. Thirdly, Chang et al. did not limit their assessment to diminutive lesions, the only situation for which a resect and discard strategy has been advocated [13]. This may have biased the results, because of differing prevalence of disease between larger and diminutive lesions (i. e., leading to different positive and negative predictive values), and because of the varying difficulty of in vivo assessment of polyps of different sizes. Poorer diagnostic performance with chromoendoscopy has been seen with decreasing lesion size [20]. Fourthly, it is debatable whether an eventual conservative three-stage strategy based on the application of white light endoscopy, ECE, and HRME at the same procedure might really be an efficient alternative to postendoscopic histological examination. Yet a fifth issue is that HRME was not sensitive enough to potentially compete with a primarily ECE-based in vivo prediction [13], being apparently condemned to play a secondary role. Finally, HRME is not the only player in the field of subcellular imaging, having competition from established confocal laser endomicroscopy [21] [22] [23]. However, HRME seems economically attractive, because of its low cost. Finally, no clinical study on HRME for colorectal polyp prediction is actually available, so that it is unknown whether this technique can satisfy officially recommended minimum criteria for incorporation of advanced endoscopic imaging into clinical practice. These have been defined by the PIVI (“Preservation and Incorporation of Valuable endoscopic Innovations”) statement of the American Society for Gastrointestinal Endoscopy (ASGE) as a 90 % agreement with post-histological surveillance recommendations and a > 90 % negative predictive value for diminutive adenomatous histology [17].

In conclusion, improvements in advanced endoscopic imaging support the adoption of a resect and discard strategy for diminutive lesions. Despite this, it is unlikely that HRME could challenge the primary role of ECE at this stage; however HRME may enrich and strengthen such strategies if it can be used to complement ECE predictions. Nevertheless, the high specificity of HRME needs to be confirmed in future in vivo clinical studies of diminutive lesions.