From January 2015 to August 2019, patients who underwent colposcopy and cervical biopsy or cervical conization in the colposcopy clinic of the Second Affiliated Hospital of Jiaxing Medical College were analyzed retrospectively. Indications for colposcopy included the following: positivity for cervical fluid-based cytology (TCT), positivity for high-risk human papillomavirus (hr-HPV), negativity for both TCT and hr-HPV, and obvious clinical symptoms: such as contact bleeding, irregular neoplasm on the cervical surface, cervical stenosis or barrel cervix. The exclusion criteria were as follows: pregnancy, previous cervical lesions or surgery, hysterectomy or incomplete data. All patients signed informed surgical consent forms before undergoing colposcopy and were informed of the purpose of the study. The study was approved by the hospital ethics committee (JXEY-2020JX009).

Colposcopy specialists have obtained the certification of colposcopy qualification and received training in these three practical techniques (IFCPC, Reid and R-way). Colposcopy was performed by 5 colposcopy specialists using R-way evaluation method, and targeted cervical biopsy was performed on abnormal areas under colposcopy. For inconclusive diagnosis, it is reviewed by another qualified senior colposcopy expert to determine the final colposcopy diagnosis.

Reid and 2011 IFCPC evaluation methods were applied for colposcopy image description and diagnosis. Each patient’s colposcopy digital photos were reviewed by trained colposcopy experts who were unaware of the patient’s previous colposcopy diagnosis. Researchers and pathologists were aware of the results of cytology or HPV testing. As shown in Fig. 1.

Fig. 1.

Flow-chart: the patient’s flow through the study.

For colposcopy, photoelectric integrated digital electronic colposcopy (Shenzhen Libang Precision Instrument Co., Ltd., Shenzhen, China) was performed by colposcopy specialists using routine 5% acetic acid tests and 5% Lugol liquid iodine staining procedure. The patient was placed in the bladder lithotomy position to expose the cervix and vaginal fornix; 0.9% sodium chloride solution was used to remove mucus secretions from the cervical surface; 5% cotton wool was used to wet the cervical surface for 50 seconds; and images were captured at 60 seconds, 90 seconds, and 120 seconds and observed within 2 minutes to assess changes after application of acetic acid and Lugol’s solution.

The R-way system evaluation method [9] is as follows: (1) The transformation zone type is evaluated. (2) The current image is captured with and without the green filter to evaluate the “red” (R), namely, the area on the surface of the cervix with the appearance of blood vessels, including physiological and abnormal new blood vessels. (3) Cotton balls containing 5% acetatic acid are used to wet the cervix for 50 seconds. Images are then, captured at 60 seconds, 90 seconds, 120 seconds, and the appearance of aceto-white epithelium is observed within 2 minutes. If the whiteness persists or thickens, this feature is defined as “white” (W). (4) Abnormal blood vessels are checked according to a rough mosaic, namely, whether large blood vessels and atypical blood vessels appear in the aceto-white area; this is referred to as “A”. (5) Lugol’s iodine solution is applied to the cervix to check iodine uptake; yellow (Y) means no iodine uptake.

The R-way evaluation system aims to diagnose HSIL+ based on the principle of collecting colposcopy examination images according to the above characteristics for correlative comparison, based on the pathological “comparison database” intelligence analysis. The four diagnostic criteria for HSIL+ evaluation with the R-way system, as shown in Fig. 2a–d are as follows: red thick white, red white A, red white yellow, and bleeding.

Fig. 2.

The diagnostic criteria of the R-way system in the evaluation of HSIL+. (a), (b), (c), and (d) indicate the diagnostic criteria of the R-way system in the evaluation of HSIL+: (a), red thick white; (b), red white A; (c), red white yellow; and (d), bleeding.

The 2011 IFCPC classification of colposcopy diagnosis is as follows: (1) In the general assessment of colposcopy images: categories include sufficient or inadequate, squamous junctions (fully visible, partially visible or invisible) and transformation zone type 1, 2 or 3. (2) Descriptions of colposcopy images include primitive squamous epithelium, columnar epithelium, metaplastic squamous epithelium, decidualization during pregnancy, and the location and size of the lesion. (3) The findings of colposcopy image grading include a thin aceto-white epithelium; small mosaic and small punctate vessels, suggesting low-grade lesions; a thick aceto-white epithelium, coarse mosaic and thick punctate vessels, with clear boundaries, suggesting high-grade lesions; lymphocytes, erosion, and negative iodine staining as non-specific changes; atypical blood vessels and other signs (exogenous lesions, necrosis, ulcers, etc.), suggestive of suspected invasive cancer changes; and miscellaneous features including condyloma acuminatum, polyps, and other obvious contact bleeding. (4) According to the above characteristics, the diagnosis of colposcopy is normal or inflammation, low-grade squamous intraepithelial lesion (LSIL), HSIL or invasive carcinoma [6,9].

The overall evaluation scores for Reid’s colposcopy range from 0–8 points, and the lesion is assigned a score from 0–2 points for each of the following features: the edge of the lesion, the colour of the acetic acid test, the shape of the blood vessel, and the results of the iodine test. Reid scores of 3–4 are usually associated with CIN1–2, and scores of 5–8 are usually associated with HSIL [7].

Regardless of the final score, biopsy forceps were used to perform colposcopy-directed biopsy and/or cervical scraping of all abnormal areas. Routinely cervical biopsies are obtained from 3, 6, 9, or 12 o’clock positions of the cervixin cases of colposcopy imaging and endocervical curettage without abnormalities, if colposcopy was unsatisfactory and squamous junctions invisible or partially visible. If colposcopy was satisfactory, squamous junctions fully visible and no abnormal lesions were found, the patient may not undergo biopsy and be excluded from the analysis. According to the 2013 American Colposcopy and Cervical Pathology (ASCCP) guidelines, the pathological diagnosis was classified as follow: normal or inflammation; LSIL; HSIL; or cervical invasive carcinoma [5].

SPSS 21.0 software (SPSS Inc., Chicago, IL, USA)was used for statistical analysis. The analysis of consistency between the colposcopy diagnosis and pathological diagnosis was performed with the Kappa test. Kappa 0.4 meant poor consistency; 0.4 $>$ Kappa (0.75 meant medium consistency; and Kappa $>$ 0.75 meant good consistency. The sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) and the Youden index were used to evaluate the authenticity and predictive value of the diagnosis. The difference between the three evaluation methods was analysed with the McNemar test. P 0.05 was considered statistically significant.

The evaluation of consistency between the colposcopy diagnosis and histopathology included a total of 987 patients, with an average age of 41.94 $\pm$ 12.45 years. The colposcopy diagnosis and pathological diagnosis of the R-way system were in agreement 69.00% (681/987) of the time, and the consistency of the test results showed a Kappa = 0.517 (P 0.001, 95% CI: 0.472–0.562). In patients with pathological diagnoses of normal cervix or inflammation, LSIL, HSIL or cervical cancer, the consistency rates of colposcopy with the pathological diagnosis were 67.43% (352/522), 68.81% (203/295), 73.13% (98/134) and 77.77% (28/36), respectively. The rate of overestimation with the HSIL+ diagnosis was 1.18% (2/170), and that of the underestimation with the LSIL+ diagnosis was 21.51% (100/465). Table 1 shows the consistency of different colposcopy diagnostic methods and cervical histopathology. The consistency rate of the IFCPC diagnosis with histopathology was 75.28% (743/987), with Kappa = 0.599 (P 0.001, 95% CI: 0.556–0.642), and the consistency rate of the Reid diagnosis with histopathology was 55.32% (546/987), with Kappa = 0.310 (P 0.001, 95% CI: 0.263–0.357). The consistency levels of the R-way and IFCPC diagnose with the pathological diagnosis were medium, while that of the Reid diagnosis was poor.

In total 170 women (17.2%) had histologically confirmed HSIL+ including 36 women (3.6%) with invasive cervical cancer. The sensitivity for HSIL+ using the three different colposcopy evaluation methods were 74.7% (127/170), 78.2% (133/170) and 62.9% (107/170) for IFCPC, R-way and Reid. Table 2 summarizes the diagnostic value of colposcopy as a predictive indicator of a lesion histological diagnosis of HSIL+ and LSIL+. When HSIL was used as the critical value, the sensitivity, specificity, PPV and NPV of IFCPC and R-way for HSIL+ were better than those of the Reid system. The sensitivity and NPV of both R-way and IFCPC were better, but IFCPC had higher specificity and PPV. When LSIL was used as the critical value, the sensitivity of each method improved, the specificity was reduced, and the Youden index was not as high as that of HSIL as the critical value.

Table 3 compares the accuracy of the three colposcopy assessment methods for HSIL+. The McNemar test results showed that the difference between R-way and IFCPC in the diagnosis of HSIL+ was statistically significant ($\chi$${}^2$ = 19.558, P 0.001); with Kappa = 0.826, the agreement rate was 94.6% (934/987). The difference between the Reid score and IFCPC for the diagnosis of HSIL+ was also statistically significant (${\chi }^2$ = 17.876, P 0.001); with Kappa = 0.127, the agreement rate was 70.62% (697/987).