Results
Participants’ characteristics and main clinical outcomes
A total of 483 patients were enrolled, among whom 71.6% were ASCUS, 24.8% were LSIL and 3.5% were ASC-H (Figure 1, Table S1). 18 patients were excluded for pathological analysis (reasons shown in Figure 1). The overall mean age was 43.5 years (range 15-71 years) with no difference among the three TCT groups. Generally, 70.6% women were non-menopausal. The overall positive rates of TS and hrHPV were 40.2%, 71.6%, respectively. Of 472 participants who completed colposcopy examination, 20.1% were cervical TZ type I, 29.2% were TZ type II and 49.8% were TZ type III. In total pathological investigations of 465 participants, there were 41.3% no CIN or worse reported, 36.1% CIN1, 11.4% CIN2, 9.9% CIN3 and 1.3% cervical cancer. Significant differences were observed among the three TCT groups regarding menstruation status, positive rate of TS and hrHPV, TZ type and pathological results (all P<0.001).
Overall performance of different screening tests for detecting CIN1+, CIN2+ and CIN3+
The sensitivity, specificity, PPV, NPV and AUC of the three screening methods (TS alone, hrHPV alone, TS in combination with hrHPV) for detecting CIN1+, CIN2+ and CIN3+ were summarized in Table S2, Table 1 and Table S3, respectively. Generally, single hrHPV test had the highest sensitivity and NPV in all settings compared with the other two methods. The overall sensitivity of TS for detecting CIN1+, CIN2+ and CIN3+ were 53.1%, 65.7%, 67.3%, respectively, while hrHPV had the corresponding significantly higher sensitivity of 82.7%, 96.2%, 96.2% (all P<0.001). The highest sensitivity of TS was 92.3% observed in ASC-H group for detecting CIN2+.
Oppositely, TS alone and TS in combination with hrHPV had significantly higher specificitiy (77.1% vs 82.8% vs 43.2% for CIN1+, 66.7% vs 73.3% vs 35.1% for CIN2+, 62.7% vs 68.4% vs 31.1% for CIN3+, all P<0.001) and PPV (76.7% vs 79.9% vs 67.4% for CIN1+, 36.5% vs 41.5% vs 30.2% for CIN2+, 18.5% vs 20.7% vs 15.0% for CIN3+, all P<0.001) than hrHPV in overall population. TS in combined with hrHPV detecting CIN1+ in ASCUS patients had the highest specificity (83.7%) in all settings.
When comparing TS alone with TS in combination with hrHPV, only in detecting overall CIN1+ patients (P<0.001) and ASCUS CIN1+ patients (P=0.004) did the two methods showed significant differences. The top three AUC were allocated to TS in combination with hrHPV in detecting overall CIN2+ patients (0.690), LSIL CIN2+ patients (0.678) and LSIL CIN3+ patients (0.677).
TS showed significantly higher specificity and NPV in incomplete cervical TZ type than TZ type I
Incomplete cervical TZ type compromises type II and III. Table 2 and Table S4 showed the general TS detecting performance in participants grouped by TZ type. For detecting CIN1+, TS demonstrated significantly higher specificity in incomplete cervical TZ type than TZ type I in ASCUS, LSIL and overall CIN1+ patients (83.1% vs 62.1%, 74.3% vs 25.0%, 81.1% vs. 57.6%, respectively. All P<0.05).
For overall CIN2+ patients, TS showed significantly higher specificity and NPV in incomplete cervical TZ type than TZ type I (70.0% vs 48.5%, 89.6% vs 27.1%, both P<0.05). Of all CIN2+ patients, TS had a higher specificity in ASCUS group (73.1% vs 53.3%, P<0.05) and a higher NPV in LSIL group (85.2% vs 26.7%, P<0.05) in incomplete cervical TZ type than TZ type I.
For detecting CIN3+, TS in both ASCUS group and LSIL group demonstrated a higher specificity in incomplete cervical TZ type than TZ type I (69.4% vs 53.0%, 63.0% vs. 23.5%, both P<0.05). In addition, CIN3+ LSIL group had a significantly higher NPV (94.4% in incomplete cervical TZ type vs 13.3% in TZ type I) and the highest AUC (0.715). In overall CIN3+ population, specificity (66.9% vs 46.4%, P<0.05), PPV (19.9% vs 6.3%, P<0.05) and NPV (95.2% vs 86.7%, P<0.05) of TS all presented higher in incomplete cervical TZ type than TZ type I.
Moreover, sensitivities of TS detecting CIN2+ and CIN3+ in all settings were all higher in incomplete cervical TZ type than TZ type I, even though no statistical significancy observed.
TS demonstrated a better clinical performance in non-menopausal patients than post-menopausal patients for detecting CIN1+
Overall detecting cases of TS in patients with TZ type III grouped by menopausal status was shown in Table S4. For detecting CIN1+, TS showed a significant difference between non-menopause group and post-menopause group (P=0.006 for ASCUS, P=0.006 for LSIL and P<0.001 for overall CIN1+ population). Moreover, the overall sensitivity for detecting CIN1+ and CIN3+ was significantly higher in non-menopausal group than post-menopausal group (45.8% vs 28.1% for CIN1+, 85.7% vs 33.3% for CIN3+, both P<0.05) (Table S5). Noteworthily, TS detecting in non-menopausal LISL group showed AUC of 0.714 (CIN1+), 0.824 (CIN2+) and 0.923 (CIN3+). For overall non-menopausal CIN2+ and CIN3+ population, the AUC were 0.736 and 0.821, respectively.
Significantly higher rate of pathology results ≤CIN1 with negative TS than positive TS results
In ASCUS patients with negative TS results, 190/213 (89.2%) patients were ≤CIN1 pathologically, while 83/117 (70.9%) ASCUS patients with positive TS results were ≤CIN1. And in LSIL patients with negative TS results, 49/61 (80.3%) patients were ≤CIN1, 34/57 (59.6%) LSIL patients with positive TS results were ≤CIN1. In ASCUS and LSIL group, both rates of pathology results ≤CIN1 with negative TS were significantly higher than which with positive TS results (P<0.001 for ASCUS, P=0.014 for LSIL) (Table S6).
Correlation between TS and IHC staining p16, Ki-67
In analysis of correlation between TS and p16, correlation was found in overall population (ρ=0.223, P<0.001) and ASCUS patients (ρ=0.226, P<0.001) (Table S7). We divided quantitative Ki-67 results into three groups for correlation analysis (<5%, 5%-30% and >30%) [22, 23]. Correlation between TS and Ki-67 were found in overall population (ρ=0.276, P<0.001), ASCUS patients (ρ=0.266, P<0.001) and LSIL patients (ρ=0.213, P=0.037).