We retrospectively analyzed data of the early esophageal cancer and precursor, which were treated with endoscopic resection (ER) at the Department of Endoscopy, Cancer Institute and Hospital, Chinese Academy of Medical Science from January 2007 to March 2011. The indications for ER in our hospital include: (1) Endoscopic and histological diagnosis of high-grade intraepithelial neoplasia (HGIN, severe dysplasia) or early esophageal cancer (ESCC); (2) low-grade intra-epithelial neoplasia (LGIN, mild dysplasia) or middle-grade intraepithelial neoplasia (MGIN, moderate dysplasia), while the diagnosis based on the endoscopic examination was worse, as HGIN or ESCC; (3) “superficial” (type 0-II) endoscopic appearance: “elevated” (type 0-IIa), “flat” (type 0-IIb) or “depressed”(type 0-IIc). The contraindications for ER included: (1) Lesions with positive lymph node metastasis (N) or distant metastasis (M); (2) EUS showing depth of invasion into the lower two-thirds of the submucosa; (3) “protruding” (type 0-I) or “excavated” (type 0-III) endoscopic appearance, according to the Paris classification; (4) esophageal varices at the range of the lesion; and (5) uncontrolled coagulopathy with international normalized ratio (INR) >2 or platelet count <75,000 per μL.
The lesions diagnosed as esophageal precursor or early cancer were stained with Lugol’s solution (if not allergic) according to the endoscopic appearance. In our study, no patients with allergy were seen, and all the lesions were stained by the Lugol’s solution. The number of the forceps biopsy was up to the length of the lesion, for instance, lesion > 1 cm corresponded with the number of the biopsy ≥ 2; the length > 2 cm, corresponded with number ≥ 3, and the length > 3 cm, the number ≥ 4. The endoscopic diagnosis was based on Lugol’s staining according to the grade of the lesions . If the diagnosis based on the endoscopic images was HGIN or ESCC (that is grade I), while the histopathological diagnosis based on the biopsy was LGIN or MGIN, ER was performed after consultation with patients. The macroscopic classification was based on Paris endoscopic classification [2, 3], classified as 0-IIa, 0-IIb, 0-IIc. We used the Paris classification to distinguish the lesions as invasive or non-invasive, to decide the appropriate therapeutic intervention. Endoscopic ultrasonography (EUS) and computed tomography (CT) were performed before ER to confirm the depth and absence of metastasis. Only lesions without metastasis were appropriate for ER. If the lesions were located in the mucosa or the upper one- third of the submucosa (sm1), the lesion was treated with ER. If the lesions invaded the lower two-third of the submucosa (sm2), surgery was indicated according to the EUS diagnosis.
The ER includes the endoscopic mucosa resection (EMR) with suction cap- and/or saline solution-assisted snare resection techniques according to Haruhiro Inouen [6–8] and multiband mucosectomy (MBM) .
Management of biopsy and ER specimens
Biopsy specimens were obtained from all unstained lesions (USLs) using standard biopsy forceps. Biopsy specimens were removed using a toothpick or other non-traumatic technique and spread out flat on a gloved finger. Each specimen was separately attached to filter paper and fixed immediately. The filter paper was later removed and the tissue was dehydrated, and embedded perpendicularly in paraffin.
Resection specimens were pinned onto the cork with the luminal side facing up. Piecemeal resection specimens were reconstructed after staining with Lugol’s solution. Specimens were subsequently fixed in 10% buffered formalin, cut into 2 mm sections, dehydrated and embedded perpendicularly in paraffin.
Slices were cut, mounted on glass slides and stained with hematoxylin and eosin. All slides of the biopsy and the ER specimens were reviewed independently by two expert gastrointestinal pathologists (L.X. and N.L.). Discordant cases were reviewed jointly until a consensus was reached. The following parameters were evaluated for invasive lesions: depth of invasion (stage), degree of differentiation (grade), lymph-vascular infiltration, submucosal invasive depth (distance from the bottom of muscularis mucosa to the base of cancer nests) .
Endoscopic diagnosis was based on the criteria established in esophageal cancer high-risk regions (Figure 1) . After staining with Lugol’s solution, the mucosa with USL turned brown, defined as negative. Weak USL with indistinct margins, defined as grade III, was usually LGIN. The weak USL with clear margin defined as grade II, was usually MGIN. HGIN or T1 lesions comprise distinct USL with clear margin, defined as grade I, with or without protruded or depressed lesions.
EUS diagnosis was based on the layers of esophagus, to detect the invasive depth of the lesion into mucosa or submucosa, for possible resection. The EUS with the Sonoprobe System was used before ER to evaluate the resected specimens histopathologically. We interpreted the depth of tumor invasion based on ultrasonography using 20-MHz probe.
The macroscopic classification was based on the Paris endoscopic classification [2, 3], classified as 0-IIa, 0-IIb, 0-IIc, to distinguish the invasive from non-invasive lesions. The Paris classification dictated the choice of the ER.
Pathological diagnosis was based on a three-tier classification of squamous IN: LGIN, confined to the lowest third of the epithelium; MGIN, the lower two-thirds and HGIN, IN including the whole epithelium.
The pre-ER diagnosis was based on the worst possible diagnostic outcome among the endoscopic examination, the Paris classification and the biopsy histopathological diagnosis. Such as, the lesion was 0-II according to the Paris classification which was the indication for the ER. If the diagnosis based on the biopsy was HGIN, while endoscopic diagnosis after the lugo’s staining as the MGIN, the pre-ER diagnosis was HGIN with the EUS no invasion to submucosa. If the diagnosis based on the biopsy was MGIN, while endoscopic diagnosis after the lugo’s staining as the HGIN/ESCC, the pre-ER diagnosis was HGIN/ESCC with the EUS no invasion to submucosa.
The gold standard for the lesion was based on a combination of biopsy and ER, the worse one between these two diagnose was the gold standard.
The under-diagnosed rate was determined, based on the pre-ER systems or the ER diagnosis compared with the gold standard. The accuracy rate was determined, based on the pre-ER systems matching the gold standard. The over-diagnosed rate was determined, based on the extent to which pre-ER systems or the ER diagnoses were over-diagnosed compared with the gold standard.
The under-diagnosis, accuracy, and over-diagnosis rate were calculated based on the number of the pre-ER as numerator and the gold standard as denominator. The under-diagnosis means the numerator was lesser than the denominator, while the accuracy means the same diagnosis, and the over-diagnosis means the numerator worse than the denominator.
The Receiver-Operating-Characteristic (ROC) curve was used to evaluate the accuracy of the pre-ER system and the ER diagnosis. The Area under each ROC (AUC) and its 95% confidence were calculated. We used the Z-test to examine the differences between the AUCs based on the pre-ER diagnosis and ER diagnosis. All statistics were performed using SPSS 17.0 for Windows.