Abstract
Background
Ovarian cancer is the most fatal tumor of the female reproductive system and the fifth leading cause of cancer death among women in the USA. The prognosis is poor due to the lack of biomarkers for treatment options.
Results
The methylation array data of 551 patients with ovarian serous cystadenocarcinoma (OSC) in The Cancer Genome Atlas (TCGA) database were assessed in this study to explore the methylation biomarkers associated with prognosis and improve the prognosis of patients. These patients were divided into training (first two thirds) and validation datasets (remaining one third). A five-DNA methylation signature was found to be significantly associated with the overall survival of patients with OSC using the Cox regression analysis in the training dataset. The Kaplan–Meier analysis showed that the five-DNA methylation signature could significantly distinguish the high- and low-risk patients in both training and validation sets. The receiver operating characteristic (ROC) analysis further confirmed that the five-DNA methylation signature exhibited high sensitivity and specificity to predict the prognostic survival of patients. Also, the five-DNA methylation signature was not only applicable in patients of different ages, stages, histologic grade, and size of residual tumor after surgery but also more accurate in predicting OSC prognosis compared with known biomarkers.
Conclusions
This five-DNA methylation signature demonstrated the potential of being a novel independent prognostic indicator and served as an important tool for guiding the clinical treatment of OSC to improve outcome prediction and management for patients. Hence, the findings of this study might have potential clinical significance.
Similar content being viewed by others
Background
Ovarian cancer is the most lethal cancer of the female reproductive system and the fifth leading cause of cancer death among women in the USA with an estimated 22,240 new cases and 14,070 deaths expected to occur in 2018 [1, 2]. Ovarian serous cystadenocarcinoma (OSC), a common type of ovarian cancer, accounts for about 90% of all ovarian cancers [2]. The standard treatment consists of cytoreductive surgery followed by a combination of platinum- and taxane-based chemotherapy [3]. Although advances in treatment technology in the last few decades have substantially improved the average survival time, the cure rates remain relatively unchanged [4]. The overall 5-year survival probability of women diagnosed with ovarian cancer is still less than 50% (47%) [1]. Assessment of patients prior to therapy might enable a risk-adapted approach and hence offer an opportunity to provide improved personalized treatment. Physicians can direct low-risk patients to conventional treatments, while high-risk patients can be channeled to trials of novel therapies. This selection may enhance the ability of clinical trials to demonstrate clinical benefits. Therefore, determining high-risk patients with OSC and improving the clinical outcome are urgently needed for current clinical management. The identification of highly specific, sensitive, and independent predictive prognostic biomarkers that will allow the stratification of care is essential.
DNA methylation is well known to be associated with ovarian cancer and has great potential to serve as a biomarker in screening the disease, monitoring response to therapy, and predicting the prognosis [Full size image
Evaluation of the predictive performance of the five-DNA methylation signature using ROC analysis
The sensitivity and specificity of the five-DNA methylation signature in predicting survival were evaluated using the ROC analysis to further assess the predictive accuracy of the five-DNA methylation signature in the validation dataset. The AUC of the five-DNA methylation signature was 0.715 (P < 0.001, 95% CI 0.62–0.81) (Fig. 2), indicating that the five-DNA methylation signature had high sensitivity and specificity. Therefore, it could be used to predict the prognostic survival of patients with OSC with high accuracy, and it might have potentially great significance in clinical application.
Predictive performance of the five-DNA methylation signature based on different regrou** methods
Furthermore, several factors were associated with prognostic survival, including age [1, 17], stage [18], histologic grade [19], and size of residual tumor after cytoreductive surgery [20], and the reproducibility was poor in the prognostic markers identified by different groups [17], and Chi et al. found that patient age might serve as a significant prognostic factor for ovarian carcinoma [21]. Patients were divided into three cohorts based on their ages at initial diagnosis: ≤ 50 (N = 127, 23.05%), 51–60 (N = 178, 32.30%), and > 60 (N = 246, 44.65%), to analyze the clinical effect of the five-DNA methylation signature in patients with different ages. Kaplan–Meier curves showed that patients in the low-risk group had significantly (P < 0.01) longer OS, and the AUC value was 0.680, 0.774, and 0.720 respectively for the three age cohorts (Fig. 3), suggesting that the five-DNA methylation signature was independent of age. Patients in stages III and IV had significantly shorter OS compared with patients in stages I and II [22], and the 5-year survival of women diagnosed with distant-stage disease was only 29% [1]. Despite the markedly different outcomes by the extent of disease, the OS was obviously different in high- and low-risk groups, and the AUC in stages I and II and stages III and IV cohorts was 0.778 and 0.735, respectively (Additional file 1: Figure S2). As for the histologic grade, considering the number of samples, we verified the predictive performance of the five-DNA methylation signature in G2 (N = 69) and G3 (N = 478). Irrespective of grades, the patients in the high-risk group had significantly (P < 0.05) shorter OS, and the AUC values were 0.696 and 0.740 (Additional file 1: Figure S3). The anatomic subdivisions from left alone and right alone were combined as unilateral cohorts for these analyses due to small numbers. The differences (P < 0.001) in the OS between the two groups were also observed, and the AUC values in all the subgroups were more than 0.72, in both unilateral (N = 140) and bilateral cohorts (N = 383) (Additional file 1: Figure S4). Recent investigations highlighted that the distribution of residual disease was an important predictor and a determinant of OS of patients [23]. The present data showed that the five-DNA methylation signature could provide a fairly better reference for different residual disease cohorts owing to the effectiveness of risk stratification (Additional file 1: Figure S5). All these results indicated that the signature showed satisfactory applicability when patients were regrouped by different clinicopathological characteristics, suggesting that the signature was an independent applicable prognostic predictor of patient survival. The results are summarized in Table 2.
Comparison of the five-DNA methylation signature with other known prognostic biomarkers
In addition, several prognostic biomarkers were identified in previous studies. For instance, Luo et al. demonstrated that the expression of HER2 was a predictor of poor prognosis for ovarian cancer [24]. The expression model of MANF combined with DOCK11 was associated with the prognostic outcomes of patients with OSC, and the model could potentially serve as a novel prognostic indicator [25]. The methylation of the BRCA1 promoter was associated with a poor patient outcome [26]. Expression of HOTAIR was an independent prognostic factor of OS, and its surrogate DNA methylation signature indicated carboplatin resistance in ovarian cancer [27, 28]. The sensitivity and specificity of known biomarkers from other studies were chosen to be evaluated in the validation dataset so as to verify whether the five-DNA methylation signature had the advantage of stable and reliable performance. The ROC analyses for other known biomarker is just as the analysis for our five-DNA methylation signature, and the results showed that the five-DNA methylation signature outperformed other known prognostic biomarkers, including the types of mRNA, lncRNA, and DNA methylation. And statistical comparison using Z test revealed that our signature had significantly higher (P < 0.05) predictive performance than most of the other known biomarker. The AUCs of these biomarkers are shown in Fig. 4 and Additional file 1: Table S2. All these results inspiringly revealed that the five-DNA methylation signature provided better stability and reliability in predicting the OS of patients with OSC and was a superior predictor. Additionally, the expression of the genes corresponding to the five DNA methylation sites and genes in the five-mRNA signature [29] whose accuracy is second only to the five-DNA methylation signature were also analyzed. And the results demonstrated that the latter genes had higher fold changes in the comparison of high- and low-risk patients, and no difference was noted in the expression of almost all the former five genes in this study (Additional file 1: Figure S6).