Narrow-band imaging in the diagnosis of deep submucosal colorectal cancers: a systematic review and meta-analysis.

Magnifying endoscopy with narrow-band imaging (M-NBI) has been widely used in the differential diagnosis of deep submucosal colorectal cancers (dSMCs) from superficial submucosal cancers (sSMCs) and intramucosal neoplasms. We aimed to pool the diagnostic efficacy of M-NBI and compare it with that of magnifying chromoendoscopy (M-CE) in diagnosing colorectal dSMC.  PubMed, EMBASE, and the Cochrane Library were searched to identify eligible studies. Meeting abstracts were also searched. A bivariate mixed-effects binary regression model was used in the meta-analysis to calculate the pooled diagnostic efficacy of M-NBI and compare it with that of M-CE in the diagnosis of dSMC. Subgroup analyses and meta-regression were conducted to explore sources of heterogeneity.  We included 17 studies: 14 full texts and 3 meeting abstracts. The pooled sensitivity, specificity, and area under the summary receiver operating characteristic curve (AUC) with 95 % confidence intervals (CIs) in diagnosing dSMC were 74 % (66 % - 81 %; I = 84.6 %), 98 % (94 % - 99 %; I = 94.4 %), and 0.91 (0.88 - 0.93), respectively, for M-NBI. The pooled sensitivity, specificity and AUC (95 %CI) were 84 % (76 % - 89 %; I = 76.9 %), 97 % (94 % - 99 %; I = 90.2 %), and 0.97 (0.95 - 0.98), respectively, for M-CE. M-NBI had lower sensitivity ( < 0.01) than M-CE with similar specificity ( = 0.32). Subgroup analyses and meta-regression indicated that endoscopic diagnostic criteria, study type, endoscope type, risk of index test bias, and histopathological diagnostic criteria might be the sources of heterogeneity.  M-NBI and M-CE had comparable specificities in diagnosing dSMC, but the sensitivity of M-NBI was slightly lower than that of M-CE.