PubMed/Medline, Embase, and the Cochrane library were systematically searched. The analysis included diagnostic studies reporting on the accuracy of endoscopic biopsy or EUS in detecting residual cancer versus complete response after nCRT for esophageal cancer with histopathology as the reference standard. Bivariate random-effects models were used to estimate pooled sensitivities and specificities and examine sources of heterogeneity.
Twenty-three studies comprising 12 endoscopic biopsy studies (1281 patients), 11 EUS studies reporting on ypT status (593 patients), and 10 EUS studies reporting on ypN status (602 patients), were included. Pooled estimates for sensitivity of endoscopic biopsy after nCRT for predicting ypT+ were 34.5% (95% confidence interval [CI], 26.0%-44.1%) and for specificity 91.0% (95% CI, 85.6%-94.5%). Pooled estimates for sensitivity of EUS after nCRT were 96.4% (95% CI, 91.7%-98.5%) and for specificity were 10.9% (95% CI, 3.5%-29.0%) for detecting ypT+, and 62.0% (95% CI, 46.0%-75.7%) and 56.7% (95% CI, 41.8%-70.5%) for detecting ypN+, respectively.
Endoscopic biopsy after nCRT is a specific but not sensitive method for detecting residual esophageal cancer. Although EUS after nCRT yields a high sensitivity, only a limited number of patients will have negative findings at EUS with still a substantial false-negative rate. Furthermore, EUS provides only moderate accuracy for detecting residual lymph node involvement. Based on these findings, these endoscopic modalities cannot be used to withhold surgical treatment in test-negative patients after nCRT. (Clinical trial registration number: CRD42015016527.)