**3. Results**

The survey was conducted between January and April 2022 with the participation of 10 Italian centers: seventy-three (n. 73) colposcopists logged in to the web platform, 56 (76.7%) of them completing the whole test, and 17 (23.3%) only partially. The mean completion rate of the test for this latter subgroup of participants was 49%. The overall number of colposcopic observations/interpretations accounted for a total of 6155, upon which the survey has been performed. According to the level of colposcopic experience and practice, 27 (37%) participants reported a < 5 year practice in colposcopy (juniors) and 46 (63%) a personal experience > 5 years (seniors). No data were available regarding the number/year of colposcopies performed by participants.

The first part of the results analysis was primarily targeted at the identification of some intrinsic features of colposcopy, with the aim of evaluating the diagnostic accuracy and QC of the second-level colposcopy-based cervical cancer prevention workup. The overall analysis of the survey data in terms of colposcopy accuracy provided sensitivity and specificity rates of 61.6% and 77.1%, respectively; according to colposcopists' experience, sensitivity was 60.6% for seniors and 62.0% for juniors, while specificity was 76.7% and 77.4%, respectively. Considering the histology threshold of CIN2+, specificity increased to 87.7% (seniors 86.2% vs. juniors 88.6%).

In details, sensitivity increased from 60.9% in low-grade cases (HPV or CIN1) to 73.7% in high-grade cases (CIN2+); no statistically significant differences were obtained comparing seniors vs. juniors' rates of sensitivity (Table 1).


**Table 1.** Diagnostic accuracy of colposcopy.

Despite lacking statistical significance, senior colposcopists sensitivity was always inferior compared to juniors, with the only exception of CIN2-CIN3 cases (64.9% vs. 62.3%); when cancer cases were added to CIN2-CIN3 in a single analysis, the sensitivity rates of the two subgroups of colposcopists were closely comparable (73.5% vs. 73.9%). As for specificity, juniors' performance was again superior.

Table 2 shows the results according to the squamocolumnar junction (SCJ) evaluation, with the adoption of the 2011 IFCPC terminology [14]. Full agreement with the experts' panel was recorded in 81.2% when a *fully visible* SCJ was present, in 51.4% in *not fully visible* SCJ cases, and in 64.9% in *not visible* SCJ cases. Comparing seniors with juniors, a significant statistical difference was observed in *not visible* SCJ cases only (67.5% vs. 60.7%; *p* = 0.011). The Cohen's *kappa* correlation coefficient accounted for 0.49 (95% CI: 0.47–0.51) when the entire group of colposcopists was considered, for 0.49 (95% CI: 0.47–0.52) in the seniors group, and for 0.48 (95% CI: 0.45–0.51) in junior colposcopists. The highest rate of incorrect SCJ interpretation was recorded within the *not fully visible* SCJ group, where it accounted for 48.6%, with no statistical difference between seniors and juniors (48.1% vs. 49.5%).



All colposcopists: *p* < 2.2−16; Cohen's *kappa* correlation coefficient = 0.49 CI 95% [0.47–0.51]. Seniors: *p* < 2.2−16; Cohen's *kappa* correlation coefficient = 0.49 CI 95% [0.47–0.52]. Juniors: *p* < 2.2−16; Cohen's *kappa* correlation coefficient = 0.48 CI 95% [0.45–0.51]. # block letters = colposcopists vs. panel full agreement; **\*** italics = incorrect SCJ judgment by colposcopists; SCJ = squamocolumnar junction; NS = not significant.

The same analysis was performed adopting the SCJ nomenclature proposal suggested by the American Society of Cervical Pathology and Colposcopy in 2017 [15], which divided the SCJ into two colposcopic categories only: *fully visible* and *not fully visible*. Full agreement with the experts increased to 75% in the *not fully visible* SCJ subgroup, with a statistically significant difference between seniors and juniors (77.1% vs. 72.8%, respectively; *p* = 0.011). The Cohen's *kappa* concordance coefficient also increased from 0.49 to 0.57 (95% CI: 0.54–0.59) for the whole set of participants, from 0.49 to 0.57 (95% CI: 0.55–0.60) for the seniors, and from 0.48 to 0.56 (95% CI: 0.52–0.59) for the juniors group. Table 3 summarizes these results.


**Table 3.** SCJ assessment (ASCCP 2017 Nomenclature [15]).

All colposcopists: *p* < 2.2−16; Cohen's *kappa* correlation coefficient = 0.57 CI 95% [0.54–0.59]. Seniors: *p* < 2.2−16; Cohen's *kappa* correlation coefficient = 0.57 CI 95% [0.55–0.60]. Juniors: *p* < 2.2−16; Cohen's *kappa* correlation coefficient = 0.56 CI 95% [0.52–0.59]. **#** block letters = colposcopists vs. panel full agreement; **\*** italics = incorrect SCJ judgment by colposcopists; SCJ = squamocolumnar junction; NS = not significant.

Table 4 shows the results regarding colposcopists' interpretation of the Transformation Zone (TZ) compared with the experts' panel.

Full agreement was observed in 73.2% of Type 1, 53.8% of Type 2, and 66.7% of Type 3 TZ cases; within each group of TZ, a statistically significant difference was demonstrated comparing seniors to juniors: in particular, Type 1 and Type 2 TZ were better identified by junior colposcopists (79% vs. 69.5% and 55.9% vs. 52.3%, respectively; *p* < 0.05), while Type 3 TZ was significantly better identified by seniors (71.7% vs. 58.3%; *p* < 0.05).

In this analysis, the highest rate of incorrect interpretation was identified in senior colposcopists evaluating Type 2 TZ cases (47.7%), while the lowest rate was recorded in juniors' evaluation of Type 1 TZ (21%).



All colposcopists: *p* < 2.2−16; Cohen's *kappa* correlation coefficient = 0.46 CI 95% [0.45–0.48]. Seniors: *p* < 2.2−16; Cohen's *kappa* correlation coefficient = 0.46 CI 95% [0.44–0.48]. Juniors: *p* < 2.2−16; Cohen's *kappa* correlation coefficient = 0.47 CI 95% [0.44–0.50]. **#** block letters = colposcopists vs. panel full agreement; **\*** italics = incorrect SCJ judgment by colposcopists; TZ = Transformation Zone.

The second part of the survey results analysis was conversely targeted to investigate the accuracy of the colposcopic procedure through the assessment of colposcopic interpretation of cervical patterns and its influence on the operators' clinical decisions.

As far as it concerned the assessment of grade (G) of the colposcopic pattern compared to proven histology, the following results were obtained: full agreement with histology was achieved in 60.59% of cases with G1/low-grade lesions, in 59.11% of G2/high-grade lesions, and in 64.64% of colposcopic patterns suspicious for cancer and histologically confirmed cervical malignancy; these concordance rates can also be seen as PPV of colposcopy.

Interestingly, 5.05% and 19.26% of cases with a histologically proven CIN2+ were categorized as colposcopically negative or G1 by participants, respectively.

On the other hand, overestimation of the colposcopic pattern reached the highest rate in histologically proven low-grade lesions (HPV-CIN1), which were classified as G2 in 24.70% of cases (Table 5).


**Table 5.** Predictive value of colposcopic grade (G).

Pearson's chi-squared test: *p* < 2.2−16; Cohen's *kappa* correlation coefficient = 0.49 CI 95% [0.47–0.51]. \* NPV; block letters = colposcopists vs. panel full agreement and PPV. G1 = minor colposcopic.

A similar analysis was performed considering the colposcopic impression formulated by colposcopists compared to histology.

A negative colposcopic impression correlated with a negative histology in 77.9% of cases, allowing this figure to be seen as NPV. Taking into consideration histologically confirmed high-grade lesions (CIN2-CIN3), which represent the main objective of the cervical cancer prevention strategy, the colposcopic impression of a high-grade lesion was correctly formulated by colposcopists in 59.4% of cases.

When cancer cases were added to CIN2/CIN3, the PPV of a high-grade lesion colposcopic impression increased to 70.5%.

The PPV of a colposcopic impression suspicious for cancer was 64.4% (*p* < 0.05; Cohen's *kappa* correlation coefficient = 0.51; 95% CI: [0.50–0.53]) (Table 6).


**Table 6.** Predictive value of colposcopic impression (CI). **Table 6.** Predictive value of colposcopic impression (CI).

*Diagnostics* **2023**, *13*, x FOR PEER REVIEW 8 of 14

*Diagnostics* **2023**, *13*, x FOR PEER REVIEW 8 of 14

Pearson's chi-squared test: *p* < 2.2−16; Cohen's *kappa* correlation coefficient = 0.51—CI 95% [0.50–0.53]. \* NPV = Negative Predictive Value; # PPV = Positive Predictive for CIN2-CIN3; <sup>≈</sup> PPV = Positive Predictive Value for cancer; LG = low-grade lesion; HG = high-grade lesion. Pearson's chi-squared test: *p* < 2.2−16 ; Cohen's *kappa* correlation coefficient = 0.51—CI 95% [0.50–0.53]. \* NPV = Negative Predictive Value; # PPV = Positive Predictive for CIN2-CIN3; [0.50–0.53]. \* NPV = Negative Predictive Value; # PPV = Positive Predictive for CIN2-CIN3; PPV = Positive Predictive Value for cancer; LG = low-grade lesion; HG = high-grade lesion.

PPV = Positive Predictive Value for cancer; LG = low-grade lesion; HG = high-grade lesion.

Directly correlated with the colposcopic impression and the G assessments, colposcopists were asked to indicate the need for taking biopsy/biopsies and the cervical site they thought was the most appropriate for histological confirmation; biopsies were performed in 3404 cases out of 6155 in the case of the experts panel (55%), and in 3482 cases out of 6155 (56%) in the case of candidates. Figures 4–6 illustrate how the biopsy/biopsies sites were indicated by colposcopists. Directly correlated with the colposcopic impression and the G assessments, colposcopists were asked to indicate the need for taking biopsy/biopsies and the cervical site they thought was the most appropriate for histological confirmation; biopsies were performed in 3404 cases out of 6155 in the case of the experts panel (55%), and in 3482 cases out of 6155 (56%) in the case of candidates. Figures 4–6 illustrate how the biopsy/biopsies sites were indicated by colposcopists. Directly correlated with the colposcopic impression and the G assessments, colposcopists were asked to indicate the need for taking biopsy/biopsies and the cervical site they thought was the most appropriate for histological confirmation; biopsies were performed in 3404 cases out of 6155 in the case of the experts panel (55%), and in 3482 cases out of 6155 (56%) in the case of candidates. Figures 4–6 illustrate how the biopsy/biopsies sites were indicated by colposcopists.

**Figure 4.** A single biopsy is indicated (correct site). **Figure 4.** A single biopsy is indicated (correct site). **Figure 4.** A single biopsy is indicated (correct site).

**Figure 5.** Multiple biopsies indicated (correct sites). **Figure 5.** Multiple biopsies indicated (correct sites).

biopsy

LG Colposcopic Impression with CIN2+ histology

**Figure 6.** A single biopsy is indicated (correct site). **Figure 6.** A single biopsy is indicated (correct site).

According to colposcopists experience, junior colposcopists performed biopsies in 52.7% of the whole set of cases, while more experienced operators performed them in 59%. Biopsies were omitted in 96.8% of cases evaluated by colposcopists as negative, in 30.4% of cases evaluated as LG lesion, in 2.1% of cases evaluated as HG lesion, and in 0.3% of cases evaluated as neoplasia. Furthermore, it was observed that as the degree of the lesion increased, the number of biopsies consistently increased; more than one single biopsy was reported in 12.6% of cases with a colposcopic impression of LG, in 52.5% of cases of HG, and in 82.5% of cases with a colposcopic impression of cancer. According to colposcopists experience, junior colposcopists performed biopsies in 52.7% of the whole set of cases, while more experienced operators performed them in 59%. Biopsies were omitted in 96.8% of cases evaluated by colposcopists as negative, in 30.4% of cases evaluated as LG lesion, in 2.1% of cases evaluated as HG lesion, and in 0.3% of cases evaluated as neoplasia. Furthermore, it was observed that as the degree of the lesion increased, the number of biopsies consistently increased; more than one single biopsy was reported in 12.6% of cases with a colposcopic impression of LG, in 52.5% of cases of HG, and in 82.5% of cases with a colposcopic impression of cancer.

The correct site for performing biopsies was recognized in 58.9%, 77.3%, and 91.7% of histologically proven LG lesions (HPV-CIN1), HG lesions (CIN2-CIN3), and cervical cancer, respectively, while an incorrect site was indicated in 16.8%, 13.6%, and 5.3%. The correct site for performing biopsies was recognized in 58.9%, 77.3%, and 91.7% of histologically proven LG lesions (HPV-CIN1), HG lesions (CIN2-CIN3), and cervical cancer, respectively, while an incorrect site was indicated in 16.8%, 13.6%, and 5.3%.

Noteworthy, non-biopsy rates accounted for 24.3% of HPV-CIN1 cases and for 12.1% of CIN2+ cases (*p* < 0.05) (Table 7). Noteworthy, non-biopsy rates accounted for 24.3% of HPV-CIN1 cases and for 12.1% of CIN2+ cases (*p* < 0.05) (Table 7).


**Table 7.** Biopsy decision. **Table 7.** Biopsy decision.

LG = low-grade lesion.

Moreover, when the analysis focused on the subgroup of cases having a CIN2+ proven histology and a colposcopic impression of LG lesion expressed by colposcopists, the correctness of biopsy performance was significantly influenced by experience: junior colposcopists had a higher non-biopsy rate (20% vs. 10.1%), while seniors had a higher rate of correctly performed biopsies (73.9% vs. 66.9%) (*p* < 0.05) (Table 8). Moreover, when the analysis focused on the subgroup of cases having a CIN2+ proven histology and a colposcopic impression of LG lesion expressed by colposcopists, the correctness of biopsy performance was significantly influenced by experience: junior colposcopists had a higher non-biopsy rate (20% vs. 10.1%), while seniors had a higher rate of correctly performed biopsies (73.9% vs. 66.9%) (*p* < 0.05) (Table 8).

**Biopsy Experience in Colposcopy**

not performed 13.6% 10.1% 20% yes, wrong site 15% 16% 13.1% yes, correct site 71.4% 73.9% 66.9%

**All Seniors Juniors**

*p* = 0.013

**Table 8.** Underestimation of colposcopic impression vs. biopsy decision.


**Table 8.** Underestimation of colposcopic impression vs. biopsy decision.

LG = low-grade lesion.
