Patient material
Of 540 women eligible for analysis, 180 had EOC, 180 had benign ovarian cyst, and 180 were healthy controls. The average age was 56 (range, 30–76) years old. The malignancies were further divided into type I (n = 70, 39%) and type II (n = 110, 61%) EOC based on histology and grade. Type I included low-grade serous (n = 16; 9%), low-grade endometrioid (n = 15; 8%), all clear cell (n = 23; 13%) and all mucinous (n = 14; 8%). Type II included high-grade serous (n = 88; 50%), high-grade endometrioid (n = 13; 7%), all undifferentiated (n = 7; 4%) and all malignant mixed mesodermal tumors (n = 2; 1%). Most patients were diagnosed at advanced stages (FIGO III+IV: 116/180, 65%). The benign cohort included endometriosis (n = 104, 58%), serous cystadenoma (n = 24, 13%), mucinous cystadenoma (n = 29, 16%), simple cystadenoma (n = 10, 6%) and mature teratoma (n = 13, 7%) (Table 1).
CircN4BP2L2 level and clinicopathologic features of type I and type II EOC
The relationship between circN4BP2L2 level and clinicopathologic features of EOC was assessed in 180 patients (Table 2). The results showed that decreased circN4BP2L2 was associated with advanced FIGO stage (p < 0.01), worse tumor grade (p = 0.03) and LNM (p < 0.01) in EOC patients. CircN4BP2L2 level had no correlation to age or histology in EOC patients. A total of 70 patients with type I tumors were enrolled to evaluate the relationship between circN4BP2L2 and clinicopathological parameters. The data showed that low expression level of circN4BP2L2 was associated with advanced FIGO stage (p < 0.01), but not correlated to age, histology, tumor grade or LNM in type I EOC. A total of 110 type II EOC patients were included to investigate the association between circN4BP2L2 level and clinicopathological features. The results revealed that decreased circN4BP2L2 was associated with LNM (p = 0.04), but not related to age, histology or tumor stage in type II EOC.
Levels of circN4BP2L2, CA125 and HE4 in type I and type II EOC
CircN4BP2L2 and CA125 could significantly separate type I from benign or normal cohort (p < 0.01), while HE4 was not significant when type I was compared to benign (p = 0.38) or normal cohort (p = 0.54). CircN4BP2L2, CA125 and HE4 were all able to separate type II from benign or normal cohort (p < 0.01). CA125 and HE4 levels were significantly different between type I and type II cohorts (p < 0.01), but not circN4BP2L2 (p = 0.91). CA125 (p < 0.01), but not circN4BP2L2 (p = 0.21) or HE4 (p = 0.99), was significant when benign cohort was compared to normal cohort (Figure 1).
The diagnostic value of circN4BP2L2 in type I EOC
Both circN4BP2L2 and HE4 could separate type I EOC from benign cohort, whereas CA125 could only distinguish late-stage type I EOC from benign cohort. Besides, circN4BP2L2, CA125 and HE4 could all separate type I EOC from normal cohort (Table 3). The median value in Type I EOC for HE4 was 61 pmol/L, and decreased to 46 pmol/L and 44 pmol/L in benign and normal cohorts, respectively. The decrease was even more notable for CA125 where median value of CA125 ranged from 85 U/ mL in type I EOC to 48 U/ mL in benign and 29 U/ mL in normal cohort. The median value in type I EOC for circN4BP2L2 was 21, and increased to 82 in both benign and normal cohort.
In discrimination between type I EOC and benign cohorts, the ROC AUC was high for circN4BP2L2 (0.84) (Figure 2a), followed by HE4 (0.70) (Figure 2c) and CA125 (0.55) (Figure 2b). The sensitivity and specificity were higher for circN4BP2L2 (69%, 87%) than those for CA125 (59%, 32%) and HE4 (54%, 76%). When comparing early-stage tumors with benign cohort, the ROC AUC was high for circN4BP2L2 (0.79 circN4BP2L2; 0.62 HE4; 0.40 CA125) (Figure 3a, Figure 3i, Figure 3e), the sensitivity was also high for circN4BP2L2 (90% circN4BP2L2; 48% CA125; 38% HE4), and the specificity was high for HE4 (76% HE4; 56% circN4BP2L2; 32% CA125). When comparing late-stage tumors with benign cohort, the ROC AUC was high for circN4BP2L2 (0.92 circN4BP2L2; 0.80 HE4; 0.75 CA125) (Figure 3c, Figure 3k, Figure 3g), the sensitivity was also high for circN4BP2L2 (87% circN4BP2L2; 77% HE4; 73% CA125), and the specificity was also high for circN4BP2L2 (94% circN4BP2L2; 76% HE4; 32% CA125).
When distinguishing type I EOC from normal cohort, the ROC AUC was high for circN4BP2L2 (0.87) (Figure 2d), followed by CA125 (0.73) (Figure 2e) and HE4 (0.70) (Figure 2f). The sensitivity and specificity were higher for circN4BP2L2 (69%, 93%) than those for HE4 (54%, 82%) and CA125 (59%, 58%). In discrimination between early-stage tumors and normal cohort, the ROC AUC was high for circN4BP2L2 (0.81 circN4BP2L2; 0.64 CA125; 0.63 HE4) (Figure 3b, Figure 3f, Figure 3j), the sensitivity was also high for circN4BP2L2 (55% circN4BP2L2; 48% CA125; 38% HE4), and the specificity was also high for circN4BP2L2 (93% circN4BP2L2; 82% HE4; 58% CA125). In discrimination between late-stage tumors and normal cohort, the ROC AUC was high for circN4BP2L2 (0.95 circN4BP2L2; 0.85 CA125; 0.80 HE4) (Figure 3d, Figure 3h, Figure 3l), the sensitivity was also high for circN4BP2L2 (87% circN4BP2L2; 77% HE4; 73% CA125), and the specificity was also high for circN4BP2L2 (98% circN4BP2L2; 82% HE4; 58% CA125).
The diagnostic value of circN4BP2L2 in type II EOC
When separating type II EOC from benign or normal cohort, statistical significance was achieved between all groups, except for CA125 in distinguishing early-stage tumors from benign cohort (Table 4). The median value for circN4BP2L2 in type II EOC was 19, and increased to 82 in both benign and normal cohort. The median value for HE4 in type II EOC was 79 pmol/L, and decreased to 46 pmol/L and 44 pmol/L in benign and normal cohorts, respectively. The decrease was even more notable for CA125 where median value of CA125 ranged from 233 U/ mL in type II EOC cohort to 48 U/ mL in benign cohort and 29 U/ mL in normal cohort.
In discrimination between type II EOC and benign cohorts, the ROC AUC was high for circN4BP2L2 (0.88) (Figure 2h), followed by HE4 (0.79) (Figure 2j) and CA125 (0.72) (Figure 2i). The sensitivity and specificity were higher for circN4BP2L2 (93%, 71%) than those for HE4 (75%, 76%) and CA125 (72%, 32%). When comparing early-stage tumors with benign cohort, the ROC AUC was high for circN4BP2L2 (0.85 circN4BP2L2; 0.77 HE4; 0.46 CA125) (Figure 4a, Figure 4i, Figure 4e), the sensitivity was also high for circN4BP2L2 (88% circN4BP2L2; 71% HE4; 46% CA125), and the specificity was high for HE4 (76% HE4; 67% circN4BP2L2; 32% CA125). When comparing late-stage tumors with benign cohort, the ROC AUC was high for circN4BP2L2 (0.88 circN4BP2L2; 0.79 CA125; 0.79 HE4) (Figure 4c, Figure 4g, Figure 4k), the sensitivity was also high for circN4BP2L2 (94% circN4BP2L2; 79% CA125; 77% HE4), and the specificity was high for HE4 (76% HE4; 69% circN4BP2L2; 32% CA125).
When distinguishing type II EOC from normal cohort, the ROC AUC was high for circN4BP2L2 (0.91) (Figure 2k), followed by CA125 (0.84) (Figure 2l) and HE4 (0.78) (Figure 2m). The sensitivity and specificity were higher for circN4BP2L2 (80%, 87%) than those for HE4 (75%, 82%) and CA125 (72%, 58%). In discrimination between early-stage tumors and normal cohort, the ROC AUC was high for circN4BP2L2 (0.89 circN4BP2L2; 0.75 HE4; 0.68 CA125) (Figure 4b, Figure 4j, Figure 4f), the sensitivity was also high for circN4BP2L2 (83% circN4BP2L2; 71% HE4; 46% CA125), and the specificity was high for HE4 (82% HE4; 77% circN4BP2L2; 58% CA125). In discrimination between late-stage tumors and normal cohort, the ROC AUC was high for circN4BP2L2 (0.92 circN4BP2L2; 0.89 CA125; 0.79 HE4) (Figure 4d, Figure 4h, Figure 4l), the sensitivity was also high for circN4BP2L2 (83% circN4BP2L2; 79% CA125; 77% HE4), and the specificity was also high for circN4BP2L2 (87% circN4BP2L2; 82% HE4; 58% CA125).