Diagnostic performance of narrowed spectrum endoscopy, autofluorescence imaging, and confocal laser endomicroscopy for optical diagnosis of colonic polyps: a meta-analysis

Summary

Background

Novel endoscopic technologies could allow optical diagnosis and resection of colonic polyps without histopathological testing. Our aim was to establish the sensitivity, specificity, and real-time negative predictive value of three types of narrowed spectrum endoscopy (narrow-band imaging [NBI], image-enhanced endoscopy [i-scan], and Fujinon intelligent chromoendoscopy [FICE]), confocal laser endomicroscopy (CLE), and autofluorescence imaging for differentiation between neoplastic and non-neoplastic colonic lesions.

Methods

We identified relevant studies through a search of Medline, Embase, PubMed, and the Cochrane Library. Clinical trials and observational studies were eligible for inclusion when the diagnostic performance of NBI, i-scan, FICE, autofluorescence imaging, or CLE had been assessed for differentiation, with histopathology as the reference standard, and for which a 2 × 2 contingency table of lesion diagnosis could be constructed. We did a random-effects bivariate meta-analysis using a non-linear mixed model approach to calculate summary estimates of sensitivity and specificity, and plotted estimates in a summary receiver-operating characteristic curve.

Findings

We included 91 studies in our analysis: 56 were of NBI, ten of i-scan, 14 of FICE, 11 of CLE, and 11 of autofluorescence imaging (more than one of the investigated modalities assessed in eight studies). For NBI, overall sensitivity was 91·0% (95% CI 88·6–93·0), specificity 85·6% (81·3–89·0), and real-time negative predictive value 82·5% (75·4–87·9). For i-scan, overall sensitivity was 89·3% (83·3–93·3), specificity 88·2% (80·3–93·2), and real-time negative predictive value 86·5% (78·0–92·1). For FICE, overall sensitivity was 91·8% (87·1–94·9), specificity 83·5% (77·2–88·3), and real-time negative predictive value 83·7% (77·5–88·4). For autofluorescence imaging, overall sensitivity was 86·7% (79·5–91·6), specificity 65·9% (50·9–78·2), and real-time negative predictive value 81·5% (54·0–94·3). For CLE, overall sensitivity was 93·3% (88·4–96·2), specificity 89·9% (81·8–94·6), and real-time negative predictive value 94·8% (86·6–98·1).

Interpretation

All endoscopic imaging techniques other than autofluorescence imaging could be used by appropriately trained endoscopists to make a reliable optical diagnosis for colonic lesions in daily practice. Further research should be focused on whether training could help to improve negative predictive values.

Funding

None.

Introduction

Colorectal cancer develops from precursor lesions called colorectal polyps, which can be detected during colonoscopy. Removal of these lesions can prevent the development of the disease.¹ Colorectal polyps can be neoplastic, adenomas, or non-neoplastic (eg, serrated polyps and inflammatory polyps). Neoplastic lesions can become malignant, but the risk of non-neoplastic lesions other than serrated polyps becoming cancerous is negligible.2, 3 Accurate in-vivo differentiation between the types of lesions would assist decision making about endoscopic treatment, especially in the distal colon, where non-neoplastic diminutive polyps (≤5 mm) can be left in situ. Additionally, such differentiation would mean that the so-called resect and discard strategy could be implemented (ie, not all lesions would need histopathological tests after removal) and decisions about appropriate surveillance intervals could be made directly after colonoscopy.⁴

In 2011, the Preservation and Incorporation of Valuable Endoscopic Innovations (PIVI) statement about real-time endoscopic assessment of the histology of diminutive colorectal polyps⁵ introduced two criteria for assessments of whether a technique or device could replace histopathological assessment (the gold standard). First, when the technology for optimum diagnosis is used to make an in-situ endoscopic diagnosis for diminutive polyps with high confidence, this technology should result in the same surveillance interval that would have been assigned after pathological assessment of polyps at least 90% of the time. Second, for a technology to be used to guide the decision to leave suspected rectosigmoid hyperplastic polyps of less than 5 mm in size in place (without resection), the technology should provide 90% or greater negative predictive value, when used with high confidence, for adenomatous histology. Practically, if an endoscopist uses a specific technology and achieves a negative predictive value of at least 90%, diminutive lesions in the rectosigmoid colon could be left in situ if they are deemed to be non-adenomatous, other lesions could be resected but not sent in for pathology, and the surveillance interval could be established immediately.⁵

In the past two decades, several new endoscopic imaging techniques have been developed to improve endoscopic differentiation between neoplastic and non-neoplastic colonic lesions beyond standard white light assessment. Narrow-band imaging (NBI, Olympus, Japan), image-enhanced endoscopy (i-scan, Pentax, Japan), and Fujinon intelligent chromoendoscopy (FICE, Fujinon, Japan)—which are also all called virtual, digital, or electronic chromoendoscopy—are all built-in endoscopic imaging techniques. NBI is a blue light technology that highlights superficial mucosal vasculature and enhances surface patterns through illumination via narrowed bandwidth filters. Both i-scan and FICE use spectral-estimation technology to reconstruct images at different wavelengths on the basis of white light images.

Other new image enhanced techniques that are widely commercially available are confocal laser endomicroscopy (CLE) and autofluorescence imaging. CLE is a system that can provide highly magnified images of gastrointestinal epithelium that are similar to histopathological images through a miniaturised confocal laser endomicroscope, either integrated into the endoscope (Pentax, Japan), or via a probe introduced down the working channel of the endoscope (Mauna Kea Technologies, France). Autofluorescence imaging makes use of differences in mucosal blood flow and endogenous fluorophores (eg, collagen, flavins, and NADPH), which change the autofluorescence signal emitted after short wavelength illumination. The signal is processed to create a false-colour image to assist differentiation between neoplastic and non-neoplastic colonic lesions.

The diagnostic performance of these techniques has been widely studied both in single studies and in single-modality meta-analyses.6, 7, 8 However, a comprehensive overview of the accuracy and precision for all available techniques has not been combined in one meta-analysis with standardised inclusion criteria, data extraction, and statistical approach. Our aim was to establish the sensitivity, specificity, and real-time negative predictive value of NBI, i-scan, FICE, autofluorescence imaging, and CLE for differentiation between neoplastic and non-neoplastic colonic lesions, with histopathology as the reference standard.