In vivo optical cellular diagnosis for uterine cervical or vaginal intraepithelial neoplasia using flexible gastrointestinal endocytoscopy -a prospective pilot study-

Patients under lumbar spinal anesthesia and sedation were placed in the lithotomy position in the operating room. After inserting a Cusco speculum into the vagina, vaginal irrigation was performed. Cell imaging in vivo was performed using endocytoscopy (H290EC) with a video processor (CV290) and light source (CLV290; Olympus Medical Systems Corp., Tokyo, Japan). This equipment provides about 500 magnification and an observation range of 570  ×  500 μm with an outer diameter of 9.7 mm [9]. The endoscope was inserted into the vagina by senior endoscopists (S. O. and K. M.) and surfaces of the cervical or vaginal lesions were observed. After washing carefully, the cells were stained using 0.1% methylene blue and endocytoscopic images of malignant areas and surrounding areas (two areas of each) were obtained. An additional movie file shows this in more detail [see Additional file 1]. Target biopsies through the endoscope were performed from each area (Radial Jaw™ 4 biopsy forceps, Boston Scientific Corporation, USA). During the procedure, we took about 5 min. After completing the endocytoscopy, scheduled procedures were performed by gynecologists.

For evaluation of cell images, two images (from the most prominent site and surrounding area) per patient were selected, and the collection process was randomized and anonymous. Reviews were performed separately by three doctors (endoscopist: M. O., gynecologists: A. N. and pathologist: M. I.). Clinical information, preoperative results of cytology and histology, and other images including images obtained by colposcopy and endoscopy were unavailable at the time of the reviews. The three reviewers classified the cell images into 5 grades according to the ECA classification published by Inoue et al. [9]. ECA 4 and 5 are indicative of endoscopic malignancy. Figure 1 shows endocytoscopic images according to the ECA classification.

Colposcopy was performed by gynecologists before this study at an outpatient clinic using traditional equipment (OCS-500; Olympus, Tokyo, Japan). Punch biopsies were taken from abnormal lesions by using Atom Biopsy Punch, A-type (ATOM Medical Corporation, Japan).

Histopathological diagnosis of biopsies and surgical materials was made according to the grades of CIN and VAIN by pathologists (Fig. 2).

The primary endpoint was diagnostic accuracy (benign or malignant) of cell imaging at the most prominent site per patient judged by the reviewers to histological diagnosis using biopsies. When the results of the three reviewers did not match, the larger one of the reviews was used for calculation of accuracy. The secondary endpoints included diagnostic accuracy of all obtained cell images to histological diagnosis obtained by biopsies and colposcopy-guided biopsies, inter-observer agreement and adverse events. Malignant lesions were defined as ECA 4 and 5 in endocytoscopic images and as CIN 3 or VAIN 3 or worse in histology.

The sensitivity of colposcopy-guided target biopsy was about 70% according to a previous report [6]. With the assumption of diagnostic accuracy using cell imaging being 90%, the number of cases achieving a power of 80% at the significance level of 2.5% (one-sided) was calculated to be 34 based on results of the binomial test (normal approximation).

JMP® Pro 11.2.0 (SAS Institute Inc.) was used for data analysis. For the primary endpoint, the Z test was performed at a significance level of 2.5% (one-sided) with respect to a threshold value of 70% (correct diagnosis rate by colposcopy-guided biopsy). Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy were calculated, and the area under the curve (AUC) from receiver operating characteristic curve (ROC) analysis was evaluated. Fleiss’ κ was used to evaluate inter-observer agreement [14]. A p value of < .05 in each analysis was considered statistically significant.

A total of 28 consecutive patients were enrolled to this study and 54 images excluding 2 images that were insufficient images were evaluated. The mean age of the patients was 43 (range, 25–68) years, and 2 patients showed local recurrences after the operation (cervical cancer and vaginal cancer). Colposcopy and biopsy were performed in 27 patients before enrolment, and CIN 3 or VAIN 3 or worse was diagnosed in 21 patients. Characteristics of the patients are shown in Table 1.

Twenty-eight endoscopic images from the most prominent sites in the patients were analyzed. Table 2 shows the relationship between the ECS classification and grade of histological dysplasia. Sensitivity, specificity, PPV, NPV and accuracy of endocytoscopic images were 95.0% (84.8–98.6%), 87.5% (61.9–96.5%), 95.0% (84.8–98.6%), 87.5% (61.9–96.5%) and 92.9% (78.2–98.0%), respectively. Endocytoscopy significantly increased the diagnostic performance compared to colposcopy-guided biopsy (P < 0.01).

Excluding 2 images with poor staining, a total of 54 images including the most prominent site and surrounding area for each patient were evaluated (Table 3). Sensitivity, specificity, PPV, NPV and accuracy of endocytoscopic images were 92.0% (79.9–97.6%), 82.8% (72.4–87.5%), 82.1% (71.4–87.1%), 92.3% (80.7–97.6%) and 87.0% (75.9–92.2%), respectively. The ROC curve is shown in Fig. 3 and the AUC was 0.89.

The relationship between colposcopy-guided biopsy and final histological diagnosis based on the operation is shown in Table 4. The accuracy of colposcopy-guided biopsy was 74.1% (64.0–84.0%), which was similar to that in previous reports. On the other hand, the accuracy of ECA classification to final diagnosis was 85.7% (73.2–85.7%).

The diagnostic performance of each observer is shown in Table 5. Inter-observer agreement among the three reviewers using Fleiss’ κ was 0.78 (0.08–9.88, P < 0.01), and indicating good agreement. There were no adverse events and endocytoscopy had no adverse effects on scheduled procedures and final diagnosis.