Diagnostic value of the 2011 International Federation for Cervical Pathology and Colposcopy Terminology in predicting cervical lesions

Objective To evaluate the diagnostic accuracy of the 2011 International Federation for Cervical Pathology and Colposcopy (IFCPC) colposcopic terminology. Methods The clinicopathological data of 2262 patients who underwent colposcopy from September 2012 to September 2016 were reviewed. The colposcopic findings, colposcopic impression, and cervical histopathology of the patients were analyzed. Correlations between variables were evaluated using cervical histopathology as the gold standard. Results Colposcopic diagnosis matched biopsy histopathology in 1482 patients (65.5%), and the weighted kappa strength of agreement was 0.480 (P<0.01). Colposcopic diagnoses more often underestimated (22.1%) than overestimated (12.3%) cervical pathology. There was no significant difference between the colposcopic diagnosis and cervical pathology agreement among the various grades of lesions (P=0.282). The sensitivity, specificity for detecting high-grade lesions/carcinoma was 71.6% and 98.0%, respectively. Multivariate analysis showed that major changes were independent factors in predicting high-grade lesion/carcinoma, whereas transformation zone, lesion size, and non-stained were not statistically related to high-grade lesion/carcinoma. Conclusions The 2011 IFCPC terminology can improve the diagnostic accuracy for all lesion severities. The categorization of major changes and minor changes is appropriate. However, colposcopic diagnosis remains unsatisfactory. Poor reproducibility of type 2 transformation zone and the significance of leukoplakia require further study.


INTRODUCTION
Cervical cancer is the most common malignancy among women in China. Because of the widespread implementation of cervical cancer screening with cervical cytology and/or human papilloma virus (HPV) testing, colposcopy has become more frequently used globally. The use of colposcopy has led to a beneficial reduction in the number of unnecessary blinded biopsies and conization procedures as well as in the frequency of cauterization therapy for cervical erosions. Although colposcopic assessment continues to advance and evolve, universally accepted colposcopic diagnostic criteria and terminology are lacking. Most investigators believe that the traditional Reid colposcopic index (RCI) [1] is simple and easy to use, and it is widely applied. www.impactjournals.com/oncotarget/ Oncotarget, 2018, Vol. 9, (No. 10), pp: 9166-9176

Research Paper
However, some scholars believe that the standard method of evaluating color, vessels, and margins of a colposcopic lesion (i.e., grading) to determine the most serious lesion for biopsy has been questioned and traditional grading systems do not accurately assess lesion severity because colposcopic impression alone is unreliable for diagnosis [2]. Most reports have shown a perfect agreement between the colposcopic impression and histology of only 32% -37%, with the weighted kappa strength of agreement of only 0.20-0. 26. The agreement of the colposcopic diagnosis and cervical pathology within one grade has been reported to be only 75% -77%, and nearly one -third of highgrade squamous intraepithelial lesions (HSIL) were under graded [3][4][5]. The goal of clinical management is to identify and treat high-grade disease to decrease the risk of developing invasive cancer. Therefore, the role of colposcopy and biopsy is to identify cervical intraepithelial neoplasia (CIN), especially HSIL. In 2011, the International Federation for Cervical Pathology and Colposcopy (IFCPC) released a new terminology system (the 2011 IFCPC version of the colposcopic terminology, following the 1975, 1990, and 2002 versions of the colposcopic terminology system) and proposed replacing all previous versions of the colposcopic terminology [6]. Although similar to the 2002 version, the new IFCPC terminology is more comprehensive and reclassified some abnormal colposcopic findings on the basis of recent research. In addition, new colposcopic criteria and descriptions of vaginal lesions were added. The aims of the present study were to 1) determine the accuracy of colposcopic diagnosis using the new IFCPC colposcopic terminology; 2) assess the significance of individual colposcopic findings in the new IFCPC colposcopic terminology system; and 3) analyze each specific colposcopic characteristic included in the new IFCPC colposcopic terminology system.

RESULTS
The data of 2262 patients were included in the analysis. The mean patient age was 41.3 years (41.3 ± 11.6 years). The agreement of the colposcopic diagnosis and cervical biopsy pathology was perfectly matched in 1482 patients (65.5%), and the weighted kappa strength of agreement was 0.480 (p < 0.01). Moreover, agreement within one grade (referring to the same grade or to a difference of only one grade) was observed in 2232 patients (98.7%), and the weighted kappa strength of agreement was 0.980 (p < 0.01). In those patients who were histologically diagnosed as normal or benign, low grade squamous intraepithelial lesions (LSIL), HSIL, or cancer, the agreement between the colposcopy diagnosis and pathological diagnosis was 63.7% (440/691), 66.4% (624/940), 67.4% (366/543), and 59.1% (52/88), respectively. There was no significant difference among the four groups (χ 2 =3.813, p = 0.282). The colposcopic diagnosis was overestimated in 279 patients (12.3%, 279/2262) and underestimated in 501 patients (22.1%, 501/2262). Within two grades (referring to the same grade or to a difference of two grades) or more, 7 patients (0.3%, 7/2262) were overestimated and 23 patients (1.0%, 23/2262) were underestimated ( Table 1). The accuracy of colposcopy for the detection of LSIL or more (LSIL+) and HSIL or more (HSIL+) is shown in Table 2.
Colposcopic findings graded as major changes had a high predictive value for HSIL/carcinoma. The sensitivity, specificity, PPV, NPV, and Youden index for the detection of HSIL+ for major changes are shown in Table 6. The two new colposcopic criteria (inner border sign and ridge sign) also had a high predictive value for HSIL+. The specificity and PPV of the inner border sign for HSIL+ was 99.6% (1624/1631) and 75.0% (21/28), respectively. The specificity and PPV of the ridge sign for HSIL+ was 99.4% (1622/1631) and 91.4% (96/105), respectively.
Of the non-specific signs, non-stained lesions were very common (85.8%, 1940/2262) and had a high sensitivity and NPV, but it was unsatisfactory for the specificity, PPV and Youden index. Although iodinestained lesions had a high specificity, PPV and NPV, the sensitivity was very low (27.3%). Among 15 patients (0.7%, 15/2262) with leukoplakia, LSIL was found in     Tables 7 and 8. On the basis of univariate analysis, factors such as TZ type (p=0.019 and p=0.001 for LSIL and HSIL+, respectively), lesion size (p=0.002 and p=0.000 for LSIL and HSIL+, respectively), colposcopic minor/major changes (both p=0.000), and no-stained lesions (both p=0.000) were all linked with the

DISCUSSION
Colposcopic examination involves the systematic evaluation of the lower genital tract. The diagnostic criteria for colposcopy continue to improve and evolve [1,6,9,10]. However, the literature regarding the accuracy of colposcopy is contradictory, and no uniform standard has been accepted. The correlation between the colposcopic impression and histology remains poor [11][12][13][14].
We analyzed the applicability of the new IFCPC colposcopic terminology for predicting cervical lesions.
The study results showed that the agreement of the colposcopic diagnosis and cervical biopsy pathology was perfectly matched in 65.5% of lesions, further, the agreement of the colposcopic diagnosis and final pathology was perfectly matched in 66.2% of lesions, with findings significantly better than those of Baum et al. and Massad et al (perfect agreement, 32-37%) [3,4]. Tatiyachonwiphut et al. [13] reported an exact agreement of 57.9%, one grade agreement of 94.1%, and a weighted kappa strength of agreement of 0.494, which were comparable to the results of our study. However, the study lacked a standard criterion or scoring system for colposcopic examination. Li et al. [14], using the 2011 IFCPC colposcopic terminology, reported findings comparable to those of our study; however, their sample size was small. Mousavi et al. and Durdi et al. [11,15] reported a higher rate of perfect agreement (77.3% and 85.8%, respectively) and a substantial level of correlation (κ = 0.74 and κ = 0.73, respectively) using the RCI scoring system, but the sample sizes were small and HR-HPV testing was not done. In addition, the RCI method depends on the severity of disease, and the predictive accuracy increases with the advancing severity of the lesion. Our study showed that there were no significant differences between the colposcopic diagnosis and cervical pathology agreement among the various grades of lesions (P = 0.282), further suggesting that the accuracy of colposcopic diagnosis using the IFCPC terminology was stable and did not vary with lesion severity. In the present study, the colposcopic diagnosis more frequently underestimated (22.1%) than overestimated (12.3%) the cervical biopsy pathology. For those patients who underwent conization or hysterectomy, although there is no colposcopic overestimate, 115 patients (33.8%, 115/340) were underestimated. An overestimated colposcopic diagnosis leads to an unnecessary cervical biopsy, whereas an underestimated diagnosis leads to missed diagnoses, possibly high-grade lesions/carcinomas. The primary task for colposcopists is to improve the accuracy of colposcopic diagnosing HSIL+ and prevent the growth of undiagnosed lesions.
In the present study, we found that the colposcopic examination has a high sensitivity(80.3%) and is capable of differentiating a healthy cervix from any cervical lesion in 80.3% of the cases; however, the specificity was low (63.7%), probably leading to over-examination and thereby increased patients cost. Some studies have reported that such over-examination causes the patient to undergo an unnecessary cervical biopsy. We also found that colposcopy had a 71.6% sensitivity for the detection of HSIL+, similar to that reported in previous studies (30%-91.3%) [9,[13][14][15][16][17]. Moreover, the specificity for detecting HSIL was high (98.0%), also comparable to that previously reported (79% -96.5%) [13][14][15]. A high specificity lowers the misdiagnosis rate, which is important with regard to clinical management because of the significant differences in the clinical management of LSIL and HSIL. The PPV of colposcopy to diagnose HSIL+ was 93.2%, which is also comparable to the findings of previous reports [9,[13][14][15]. The NPV of HSIL was 89.9%, again comparable with previous studies (83.8%-93.5%) [13][14][15]. On the basis of these findings, if a cervical biopsy specimen was obtained from every abnormal area detected during colposcopy, 28.4% of patients would benefit from early detection and treatment of HSIL+, and only 2.0% of patients would have undergone an unnecessary cervical biopsy.
Precise identification of the squamocolumnar junction and evaluation of the TZ are crucial steps during colposcopy. The 2011 IFCPC colposcopic terminology confirmed the classification of the TZ as an obligatory terminology. Our study showed a linear trend between the TZ type and three age groups, with type 3 TZ found more commonly in women > 50 years old and type 1 TZ found more frequently in women < 30 years old. Further, in the present study, univariate analysis revealed that TZ type was linked to the diagnostic accuracy of colposcopy. However, binary logistic regression analysis revealed that TZ type was not an independent predictor factor in diagnosing LSIL or HSIL/carcinoma. A prospective multicenter study by Luyten et al. compared the distribution of TZ types among women according to age [18]. In their study, type 3 TZ was most commonly reported in women >50 years old (70%). However, there was evidence of a heterogeneous distribution of TZ types 1 and 2 between the age groups, possibly because of a diagnostic uncertainty between these types. They proposed a more precise anatomic distinction between types 1 and 2 in the IFCPC terminology to improve the utility of the new IFCPC classification and to encourage standardized universal reporting of findings. However, the 2017 American Society for Colposcopy and Cervical Pathology (ASCCP) colposcopy standards [19] did not divide cervical TZ into types 1, 2, and 3, mainly because of the poor reproducibility of type 2 TZ in actual clinical work. With regard to cervical TZ, more clinical studies are needed to confirm its feasibility and clinical value in colposcopic examination.
Lesion size has also been included in the new terminology system. Our study revealed a linear trend between lesion size and severity of the diagnosis, and a larger lesion size was more closely related to HSIL+. However, binary logistic regression analysis revealed that lesion size was not an independent predictor factor in diagnosing HSIL+. Several previous studies have shown that a large lesion size was more closely related to HSIL or more, and a small lesion size predicted the absence of a squamous epithelial lesion/CIN [17,20,21].
The 2011 IFCPC terminology has revised the grading nomenclature by adding two new pathognomonic, objective colposcopic criteria: the "inner border sign" and "ridge sign". We analyzed these specific signs in detail, and the specificity and PPV of both signs were high. Other studies have confirmed the clinical utility of these criteria and their strong association with HSIL [22][23][24][25]. With regard to other minor changes and major changes, the incidence of a thin acetowhite epithelium was the highest (65.4%). However, a thin acetowhite epithelium was found in only 27.1% of cases with HSIL, and had little or no association with HSIL+. Some studies have reported that a thin acetowhite epithelium does not suggest the existence of a cervical vaginal lesion [17]; however, other studies have reported a thin acetowhite epithelium with HSIL in an atrophic area [26]. Massad et al. proposed that all acetowhite lesions should be biopsied to maximize the sensitivity of colposcopic diagnosis while maintaining a good specificity [16]. In the present study, we found that two or more colposcopic minor changes were independent predictor factors in diagnosing LSIL. Additionally, we also found that a dense aceto-white epithelium, coarse mosaic, and coarse punctation had a high specificity and PPV for HSIL+ and were significantly associated with severe cervical lesions, and two or more colposcopic major changes were independent predictor factors in diagnosing HSIL+. These findings further confirm that the categorization of major changes and minor changes is appropriate. www.impactjournals.com/oncotarget Another common sign, a non-stained lesion, was once classified as a major change in the RCI scoring system and previous IFCPC colposcopic terminology. However, non-stained lesions are currently classified as a non-specific sign in the 2011 IFCPC terminology. Our study further confirmed that the diagnostic value of nonstained lesions for HSIL+ is very low, and was not an independent predictor factor for HSIL+; although iodinestained lesions had a high specificity, PPV, and NPV, the sensitivity was very low. Leukoplakia had a considerable PPV of 53.3% for HSIL+; however, our data revealed that leukoplakia was not an independent predictor factor for HSIL+. The present published studies are discordant on the significance of this terminology. In addition, it is worth noting that 30 of 37 cases suggestive of condyloma were related to LSIL/HSIL, suggesting that we should pay close attention to cervical condyloma during colposcopy, and that cervical biopsy should be performed if indicated.
High risk HPV-related infections are a significant source of cervical morbidity and mortality. Our data revealed that HPV type 16 was identified most frequently, followed by HPV types 52, 58, 56, 33, and 18. The prevalence of HPV type 16 increased with the grade of the cervical disease as defined by biopsy-histological diagnosis, and the presence of HPV type 16 was an independent factor for predicting cervical HSIL+ development, whereas the presence of HPV types 18, 33, 52, 56, and 58 were not associated with this risk. This implies that HR-HPV infections, especially HPV type 16, increase the risk of developing severe lesions. As colposcopy is inadequate for patients or in the case of suspected cervical HSIL+, colposcopic examination should be combined with cervical cytology and HR-HPV genotyping, especially with HPV type 16, to further improve diagnostic accuracy.
Our study had several limitations. First, the study design was retrospective in nature and enrollment bias may have occurred. Second, colposcopy diagnosis may have been biased by prior knowledge of the cervical cytology results. Third, although the conization/hysterectomy data had been used to strengthen the pathology diagnosis, the sample size is less. Finally, we did not compare the overall diagnostic accuracy of 2011 IFCPC criteria with that of other colposcopic diagnostic criteria (such as RCI) for the detection of high-grade CIN.
The clinical application of the 2011 IFCPC terminology is in its preliminary stage. Our data revealed that the categorization of major changes and minor changes is appropriate; the accuracy of colposcopic diagnosis using the IFCPC terminology was stable and did not vary with lesion severity. However, there was only a moderate strength of agreement between colposcopic diagnosis and cervical pathology, and the diagnostic accuracy remains unsatisfactory. Therefore, we propose that colposcopic findings should be combined with cervical cytology and HR-HPV genotyping test, especially with HPV type 16, to further improve the diagnostic accuracy. Studies evaluating the 2011 IFCPC colposcopic terminology are limited and further prospective studies are needed. In future studies, we aim to investigate RCI and 2011 IFCPC criteria and obtain data with respect to accuracy of the two colposcopic diagnostic methods.

Subjects
We retrospectively reviewed the medical records of women who underwent colposcopic examinations and cervical biopsy at the colposcopy clinic of the General Hospital of Tianjin Medical University between September 2012 and September 2016. These women had no previous surgical procedures of the cervix, no history of pelvic radiation or hysterectomy, no sexual intercourse for 3 days before the examination, and no confirmed or clinically suspected immunosuppression (e.g., HIV, corticosteroids, chemotherapy) or other chronic disease that might have compromised their immune system. Colposcopy indications included atypical or abnormal cervical cytology results and/ or positivity for HR-HPV and/or suspicious clinical findings. Positive cervical cytology included atypical squamous cells of uncertain significance; atypical squamous cell, cannot exclude HSIL; LSIL; HSIL; atypical glandular cells; and invasive cervical cancer. HPV genotyping was performed via the PCR-reverse dot hybridization method. Eighteen HR-HPV types (16,18,31,33,35,39,45, 51, 52, 53, 56, 58, 59, 66, 68, 73, 82, and 83) and five LR-HPV types (6,11,42,43, and 81) were identified. The data in this study were collected from the hospital's archived database. This study did not influence the diagnosis or treatment of the patients.

Colposcopy and biopsy procedure
At the colposcopy clinic, all colposcopies were performed by two expert colposcopists with more than five years of experience who had received the ASCCP professional colposcopy training. All colposcopic examinations were performed and recorded using the VIZ-YD system, which is a video exoscope-based system (optical electronic integration colposcopy, Beijing SWSY technology Co., Ltd, China) that allows full HD video documentation of the colposcopic examination and is used for clinical records and research. Steps in colposcopic assessment of the cervix included the following: 1) clean the cervix with normal saline; 2) assess the cervix about 1 min after the application of 5% acetic acid, and observe margins of the lesion, epithelial color and vascular patterns; and 3) assess the cervix after the application of diluted Lugol's iodine solution, and observe iodine staining. Colposcopic findings were described using the 2011 IFCPC terminology: (1) TZ types, lesion size in percentage of cervix, and normal colposcopic findings (such as original squamous epithelium: mature or atrophic, columnar epithelium ectopy, Nabothian cysts, crypt [gland] openings); (2) abnormal colposcopic findings, grade 1( minor) (thin acetowhite epithelium[irregular, geographic, border], fine punctation, fine mosaic), grade 2 (major) (dense acetowhite epithelium, rapid appearance of acetowhitening, cuffed crypt [gland] openings, coarse punctation, coarse mosaic, sharp border, inner border sign, ridge sign), non specific (Leukoplakia [keratosis, hyperkeratosis], Erosion, stained/non-stained Lugol's staining [Schiller's test]); (3) suspicious for invasion, atypical vessels, additional signs; and (4) miscellaneous findings, such as condyloma, polyp, inflammation, and post treatment consequence. It is impossible that all abnormal findings (suggestive of the same grade disease) simultaneously exist in one patient. As a result, colposcopists made a preliminary colposcopic diagnosis for each colposcopic finding on the basis of the criteria of the IFCPC. It should be noted that the same colposcopic finding could be present in different grades of lesions. When doubtful diagnosis occurred, all the inconsistent diagnoses were analyzed by another senior colposcopist, who made a decision on the final colposcopic diagnosis.
All colposcopically detected abnormal areas were biopsied, and large or multiple lesions were subjected to more than one biopsy. If the colposcopic examination showed no lesion in the cervix, a random biopsy specimen was obtained at the squamocolumnar junction in that quadrant at 3, 6, 9, or 12 o'clock. An endocervical curettage was performed after the cervical biopsy.
All biopsy specimens were fixed in formaldehyde and embedded in paraffin according to the routine procedures of the pathology department. The biopsy specimens were examined by the pathologist. After examining the specimen, an experienced expert pathologist made a final diagnosis according to the 2012 Lower Anogenital Squamous Project's two-tier squamous intraepithelial lesions (SIL) terminology (LSIL and HSIL). In uncertain cases, P16 immunohistochemistry should be used in the assessment of HSILs, particularly in dealing with a specimen that already has an H&E morphologic interpretation of HSIL(CIN 2) and in the distinction between HSIL(CIN2 or CIN3) and its mimics, such as immature squamous metaplasia, atrophy, reparative epithelial changes, and tangential cutting [7]. LSIL refers to koilocytosis, flat condyloma, CIN 1, or P16 negative CIN2; HSIL refers to CIN3 or P16 positive CIN2 [8]. A pathology slide review was not performed. SPSS 20.0 software (SPSS, Inc., Chicago, IL) was used for statistical analysis. The significance of agreement between the colposcopic diagnosis and cervical pathology was assessed with the chi-square statistic (χ 2 test) and weighted kappa values. The sensitivity, specificity, PPV, NPV, and Youden index were used to compare the colposcopic diagnosis and cervical pathology. P-values < 0.05 were considered statistically significant.

Author contributions
Aiping Fan: data collection, manuscript writing, project development; Chen Wang: data collection, manuscript writing; Liqin Zhang: data collection and analysis; Ye Yan: data analysis; Cha Han: data collection; Fengxia Xue: project development, manuscript revision.