1. Comparison of Septin9 methylation in cervical scrapings with different severity of cervical lesions
In fresh tissue specimens, the methylation rates of Septin9 in cervical cancer tissues (8/8) were significantly higher than those for cervicitis tissues (0/9) (100% vs 0%, P < 0.0001).
In cervical scraping samples, the Cochran-Armitage trend test showed there was a linear trend between cervical lesions and Septin9 methylation. We found Septin9 showed a clear tendency to increase methylation frequency according to the grade of cervical lesion (Cochran-Armitage trend test, P < 0.0001), with 14.81% (8/54), 25.00% (13/52), 48.00% (36/75), and 95.52% (64/67) in the no-CIN, LSIL, HSIL, and cervical cancer groups, respectively (Table 1).
Table 1
Comparison of Septin9 methylation with different severity of cervical lesions
Histological category | Methylation scores median () | Frequency of Septin9 methylation (%) | χ2 | P * |
Cervical cancer | 14.07 (5.52–29.32) | 95.52 (64/67) | | |
HSIL a | 1.77 × 10− 2 (4.00 × 10− 4-3.33) | 48.00 (36/75) | 95.31 | 1.585 × 10− 20 |
LSIL b,d | 1.40 × 10− 3 (3.00 × 10− 4-1.52 × 10− 2) | 25.00 (13/52) | | |
No-CIN c,e,f | 1.00 × 10− 3 (4.00 × 10− 4-6.20 × 10− 3) | 14.81 (8/54) | | |
Septin9 methylation scores were calculated using the following formula: 2[Ct (ACTB) -Ct (Septin9)] × 100. |
Frequency of Septin9 methylation were used to calculate the statistical difference: |
P*<0.0001, Pearson Chi-square Test |
a: compared with cervical cancer, χ2 = 38.37, P < 0.0001, Pearson Chi-square Test |
b: compared with cervical cancer, χ2 = 63.80, P < 0.0001, Pearson Chi-square Test |
c: compared with cervical cancer, χ2 = 80.83, P < 0.0001, Pearson Chi-square Test |
d: compared with HSIL, χ2 = 6.86, P༞0.05, Pearson Chi-square Test |
e: compared with HSIL, χ2 = 15.38, P < 0.0001, Pearson Chi-square Test |
f: compared with LSIL, χ2 = 1.17, P༞0.05, Pearson Chi-square Test |
P = percentiles |
The median methylation scores of the four groups tested were 1.00 × 10− 3, 1.40 × 10− 3, 1.77 × 10− 2, and 14.07, in no-CIN, LSIL, HSIL, and cervical cancer samples, respectively, with significant statistical differences across the four groups (P < 0.0001). In pairwise comparisons between groups, the median methylation scores of Septin9 in the HSIL, LSIL, and no-CIN groups were all lower when compared with the cervical cancer group (P < 0.0001); the median methylation scores of the HSIL group were also higher than those of the no-CIN group (P < 0.05). However, there was no difference in methylation levels between the LSIL group and no-CIN controls, or between LSIL and HSIL groups (P > 0.05) (Table 1; Fig. 3).
2. Clinical performance indicators of Septin9 methylation, cytology (threshold ≥ ASC-US), and HPV16/18 genotyping for discrimination of cervical cancer and ≥ HSIL.
With respect to detecting cervical cancer, Septin9 methylation had a similiar sensitivity of 95.52% with any of the other testing methods, retaining the highest specificity of 68.51%. ROC curve analysis manifested the optimal AUC of 0.92 for Septin9 methylation in the detection of cervical cancer, with satisfactory NPV, PPV, and YI. (Fig. 4A; Table 2)
Table 2
Power of Septin9 methylation, cytology and HPV16/18 genotypin for detection ≥ HSIL and cervical cancer
Triage | Endpoint | AUC (95%CI) | Sensitivity(%) (95%CI) | Specificity(%) (95%CI) | PPV(%) (95%CI) | NPV(%) (95%CI) | YI (%) |
Septin9 methylation | ≥HSIL | 0.80 (0.74–0.86) | 70.42 (0.63–0.78) | 80.19 (0.73–0.88) | 82.64 (0.75–0.88) | 66.93 (0.58–0.75) | 50.61 |
Cervical cancer | 0.92 (0.89–0.96) | 95.52 (0.91-1.00) | 68.51 (0.62–0.75) | 52.89 (0.44–0.62) | 97.64 (0.93–0.99) | 64.03 |
Cytology (≥ ASC-US) | ≥HSIL | 0.71 a (0.64–0.78) | 90.14 c (0.85–0.95) | 51.89 d (0.42–0.61) | 71.51 (0.65–0.78) | 79.71 (0.69–0.88) | 42.03 |
Cervical cancer | 0.67 b (0.60–0.74) | 97.01 * (0.93–1.01) | 37.02 e (0.30–0.44) | 36.31 (0.30–0.44) | 97.10 (0.90–0.99) | 34.03 |
HPV16/18 genotyping | ≥HSIL | 0.68 a (0.61–0.75) | 62.68 * (0.55–0.71) | 73.58 ** (0.65–0.82) | 76.07 (0.68–0.83) | 59.54 (50.98–67.56) | 36.26 |
Cervical cancer | 0.74 b (0.67–0.81) | 82.09 * (72.91–91.27) | 65.75 ** (0.59–0.73) | 47.01 (0.38–0.56) | 90.84 (0.85–0.95) | 47.83 |
HPV16/18 or Septin9 methylation | ≥HSIL | 0.71 a (0.65–0.78) | 82.39 c (0.76–0.89) | 60.38 d (0.51–0.70) | 73.58 (0.66–0.80) | 71.91 (0.62–0.80) | 42.77 |
Cervical cancer | 0.74 b (0.68–0.80) | 98.51 * (0.96–1.01) | 48.62 e (0.41–0.56) | 41.51 (0.34–0.49) | 98.88 (0.94-1.00) | 47.13 |
AUC = area under curve; YI = youden index |
* compared with the sensitivity of Septin9 methylation in group of ≥ HSIL and cervical cancer, P༞0.05 |
** compared with the specificity of Septin9 methylation in group of ≥ HSIL and cervical cancer, P༞0.05 |
a: compared with the AUC of Septin9 methylation in group of ≥ HSIL, P < 0.05 |
b: compared with the AUC of Septin9 methylation in group of cervical cancer, P༜0.05 |
c: compared with the sensitivity of Septin9 methylation in group of ≥ HSIL, P༜0.05 |
d: compared with the specificity of Septin9 methylation in group of ≥ HSIL, P༜0.05 |
e: compared with the specificity of Septin9 methylation in group of cervical cancer, P༜0.05 |
Regarding the detection of ≥ HSIL, although the sensitivity of Septin9 methylation analysis was slightly lower relative to cytologic testing (70.42% vs 90.14%, P < 0.05), the specificity was significantly higher (80.19% vs 51.89%, P < 0.05). Testing for methylated Septin9 also had a relatively lower sensitivity but higher specificity than combined tests for Septin9 methylation with HPV16/18 genotyping (P < 0.05), for distinguishing ≥ HSIL from ≤ LSIL. In addition, analysis of Septin9 methylation demonstrated a satisfactory PPV and YI in the detection of ≥ HSIL than any other test. ROC curve analysis revealed that Septin9 methylation exhibited a strong diagnostic accuracy in discriminating ≥ HSIL from ≤ LSIL, with the highest AUC of 0.80, and a sensitivity of 70.42% and specificity of 80.19%. (Fig. 4B; Table 2)
3. Relationship between pelvic nodal metastasis of cervical cancer and Septin9 methylation in ctDNA from cervical cancer plasma
Of the 113 cervical cancer specimens, 32 patients were positive for pelvic lymph nodes (LN(+)) and 81 were negative for pelvic lymph nodes (LN(-)) as confirmed by postoperative histopathology. The average age of the 113 patients with cervical cancer was 48.08 years; and neither age, pathological type, nor tumor size affected the status of pelvic nodal metastasis (P > 0.05). However, cervical cancer stage was associated with pelvic nodal metastasis, and locally advanced cervical cancer (IB2 or above) was more prone to metastasis (P < 0.05). (Table 3)
Table 3
Relationship between clinicopathological features of 113 cervical cancer and status of pelvic lymph node metastasis
Characteristics | n/N | Frequency of pelvic LN metastasis (%) | χ2 | P |
Age (average 48.0796 ± 10.009) | | | | |
≤ 48 | 12/48 | 25.00 | 0.45 | 0.501 * |
༞48 | 20/65 | 30.77 | | |
Histology | | | | |
SCC | 28/99 | 28.28 | 0.00 | 1.000 ** |
Non-SCC | 4/14 | 28.57 | | |
Stage | | | | |
IA or IB1 | 12/62 | 19.35 | 5.44 | 0.020 *** |
IB2 or above | 20/51 | 39.22 | | |
Size | | | | |
≤ 4 | 21/84 | 25.00 | 1.78 | 0.183 * |
༞4 | 11/29 | 37.93 | | |
n: the number of patients with pelvic lymph node metastasis; N: the number of total patients |
SCC = squamous cell carcinoma |
* Pearson Chi-square Test, P༞0.05; |
** Correction for continuity Test, P༞0.05; |
*** Pearson Chi-square Test, P༜0.05. |
We first defined > 1.5 ng/ml as positive for SCC-Ag based on the laboratory of the Guangdong Children and Women Hospital. Our study found no significant difference in the positive rates of SCC-Ag (threshold, 1.5 ng/ml) between the LN(+) and LN(-) groups (65.62% vs 53.09%, P > 0.05). However, the Septin9 methylation frequency in the LN(+) group was higher than in the LN(-) group (50% vs 18.52%, P < 0.05). (Table 4)
Table 4
Relationship between status of pelvic nodal metastasis and Septin9 methylation in ctDNA from 113 cervical cancer plasma
Clinical Parameters | LN(+) | LN(-) | P |
Frequency of Septin9 methylation (%) | 50.00 (16/32) | 18.52 (15/81) | 0.001 a |
Frequency of SCC-Ag (%) (threshold 1.5 ng/ml) | 65.62(21/32) | 53.09(43/81) | 0.226 b |
SD = Standard Deviation; P = Percentiles; SCC-Ag = squamous cell carcinoma antigen; LN (+) = positive for pelvic lymph node; LN(-) = negative for pelvic lymph node |
a: Pearson Chi-squared Test, χ2 = 11.42, P༜0.05; |
b: Pearson Chi-squared Test, χ2 = 1.47, P༞0.05. |
When we evaluated the effects of Septin9 methylation and SCC-Ag on the status of pelvic nodal metastasis by logistic regression analysis, we found that the regression model was statistically significant. When two independent variables, such as SCC-Ag (positive or negative) and Septin9 methylation (yes or no) included, only Septin9 methylation status was found to be meaningful for the prediction of pelvic nodal metastasis. The OR value suggested that the risk of pelvic nodal metastasis in patients with methylated Septin9 was 4.193 times (95% CI, 2.27–12.85) than that without methylation (P < 0.05).
Septin9 methylation possessed a sensitivity equal to that of SCC-Ag in predicting pelvic nodal metastasis (50.00% vs 65.63%, P > 0.05), but a higher specificity (81.48% vs 46.91%, P < 0.05). A co-test of methylated Septin9 with SCC-Ag was utilized to improve the sensitivity in differentiating LN(+) patients from LN(-) patients, and we acquired an increasing sensitivity from 50% to nearly 80%. In addition, Septin9 methylation was also equivalent to an AUC of 0.66 when comparing all the three testing (P > 0.05). (Table 5)
Table 5
Power of Septin9 methylation in plasma and SCC-Ag for prediction of pelvic nodal metastasis
Category | AUC (95%CI) | Sensitivity (%) (95% CI) | Specificity (%) (95% CI) | PPV (%) (95% CI) | NPV (%) (95% CI) |
Septin9 methylation (yes or no) | 0.66 (0.54–0.78) | 50.00 (0.33–0.67) | 81.48 (0.73–0.90) | 51.61 (0.35–0.68) | 80.49 (0.71–0.88) |
SCC-Ag (positive or negative) | 0.56 * (0.45–0.68) | 65.63 ** (0.49–0.82) | 46.91**** (0.36–0.58) | 32.81 (0.23–0.45) | 77.55 (0.64–0.87) |
SCC-Ag or Septin9 methylation | 0.61 * (0.50–0.72) | 78.13 *** (0.64–0.92) | 43.21 **** (0.32–0.54) | 35.21 (0.25–0.47) | 83.33 (0.70–0.92) |
*: compared with Septin9 methylation, P༞0.05 |
**: compared with the sensitivity of Septin9 methylation, P༞0.05 |
***: compared with the sensitivity of Septin9 methylation, P༜0.05 |
****: compared with the specificity of Septin9 methylation, P༜0.05 |