Background

Ovarian cancer (OC) is the second most common gynecologic cancer after cervical cancer and the most lethal gynecologic cancer [1]. Based on cancer statistics, approximately 22,280 new cases are estimated in 2016 in the USA, leading to 14,240 deaths due to OC [1]. Serous carcinoma is the most common histotype of ovarian cancer and accounts for the majority of deaths [2]. This is possibly due to the lack of symptoms of early-stage disease and effective early detection methods; less than 20 % of ovarian cancer patients can be diagnosed early [3]. The diagnosis of ovarian cancer is frequently determined at an advanced stage and most patients are treated via surgery combined with chemotherapy drugs [4]. The 5-year relative survival rate of ovarian cancer patients is only 38 % [1].

Studies have proven that epigenetic modifications, including DNA methylation, histone modifications, nucleosome positioning and non-coding RNAs, are early and frequent events in cancer [5, 6]. DNA methylation as a major mechanism of epigenetic modifications plays an important role in carcinogenesis and cancer progression [7, 8]. Cancer is a genetic disease that involves abnormalities of oncogenes and/or tumor suppressor genes (TSGs) [9]. Aberrant DNA methylation of CpG islands in the promoter regions leads to the silencing of tumor suppressor genes in cancer [10, 11]. The adenomatous polyposis coli (APC) gene is a tumor suppressor gene located on chromosome 5q21 that encodes a large multidomain protein [12]. The dysfunction of the adenomatous polyposis coli (APC) protein participates in tumorigenesis [13]. The APC gene has a major role in WNT signaling, cell cycle regulation, cell differentiation and proliferation, transcriptional activation, chromosomal instability, and apoptosis [14–16].

However, an individual study with a small number of subjects may lack strong statistical power. Thus, we systematically investigated studies of APC promoter hypermethylation and OC to evaluate the correlation between APC promoter hypermethylation and OC. Moreover, we also validated the clinicopathological significance of hypermethylated APC in ovarian cancer.

Methods

Literature search

The relevant literature were found in the PubMed, Embase, EBSCO, Cochrane Library, CNKI and Wanfang databases without language limitations. The following keywords or search terms were used: (adenomatous polyposis coli OR APC) AND (ovarian OR ovary) AND (cancer OR carcinoma OR tumor OR neoplasm) AND (methylation OR epigene*), updated to March 21st, 2016.

Selection criteria

The eligible studies had to meet the following inclusion criteria: 1) cancer patients were diagnosed as primary ovarian carcinoma by histopathological examination; 2) studies were associated with APC gene promoter methylation and ovarian cancer; 3) the methylated APC gene must have sufficient data about the frequencies of promoter methylation to evaluate the correlation of APC promoter methylation and ovarian cancer with clinicopathological features; 4) if the authors used duplicate sample data and published more than one paper, only the most recent paper or the most complete paper with the larger sample size was applied. The excluded studies did not meet the above inclusion criteria.

Data extraction

For the eligible studies included in the current meta-analysis, the relevant information were extracted as follows: the first author’s surname, publication year, country, ethnic population, methylation detection method, sample type, the number of APC promoter methylation events, sample size, and clinicopathological characteristics, such as tumor grade, tumor stage and tumor histology. Benign lesions, normal samples and low malignant potential (LMP) tumors were defined as controls. Moreover, tumor grades of ≤ 2 were defined as low-grade, and a tumor grade of 3 was defined as high-grade. Tumor stages of ≤ 2 were defined as early stage, while tumor stages of 3–4 were defined as advanced stage. Non-serous histotypes consisted of clear cell carcinoma (CC), endometrioid carcinoma (EC), mucinous carcinoma (MC) or transitional cell carcinoma (TC).

Statistical analysis

The present study was performed using STATA software (version 12.0, Stata Corporation, College Station, TX, USA). The pooled odds ratio (OR) and corresponding 95 % confidence interval (95 % CI) were calculated and summarized to assess the relationship between APC promoter methylation and ovarian cancer. Heterogeneity of eligible studies was evaluated based on Cochran’s Q test and I2 statistic [17]. If I2 ≥ 50 % and p < 0.1, significant heterogeneity was observed and a random-effects model was used; otherwise, a fixed-effects model was applied, indicating a lack of heterogeneity [18, 19]. Publication bias was detected using Egger’s linear regression test [20].

Results

The characteristics of included studies

Initially, we searched 94 potentially relevant articles using the above databases and keywords. According to the inclusion criteria, the final 12 studies that met the selection criteria were included in the present meta-analysis (Fig. 1). The methylation region was the promoter. Among the 12 studies, only 1 study used quantitative methylation specific PCR (QMSP) detection; the others used the methylation specific PCR (MSP) test. Twelve studies assessed the correlation between APC promoter methylation and ovarian cancer, including 10 cancer-normal studies, 7 cancer-benign studies and 5 cancer-LMP studies. In addition, 8 studies evaluated the association between APC promoter methylation and clinicopathological features, including 4 studies of tumor grade, 5 studies of tumor stage and 7 studies of tumor histology. The basic characteristics of selected studies were presented in Table 1 [21–32].

Fig. 1
figure 1

Flow diagram of the literature search strategy to identify studies

Full size image
Table 1 The basic characteristics of the included studies
Full size table

The association of APC promoter hypermethylation and ovarian cancer

When cancer patients were compared to normal samples, benign lesions, and LMP patients, the fixed-effects model was used in the current meta-analysis, indicating a lack of heterogeneity (I2 = 0.0 %, P = 0.679; I2 = 0.0 %, P = 0.938; I2 = 0.0 %, P = 0.663; respectively) (Table 2).

Table 2 Summary of the pooled OR
Full size table

The results showed that the pooled OR of the APC promoter hypermethylation was significantly higher in ovarian cancer than in normal samples and benign lesions (OR = 6.18, 95 % CI = 4.02–9.51, P < 0.001; OR = 3.26, 95 % CI = 1.65–6.44, P = 0.001; respectively), including 10 studies of 706 ovarian cancer patients and 429 normal samples and 7 studies of 431 ovarian cancer patients and 109 benign lesions (Figs. 2 and 3).

Fig. 2
figure 2

Forest plot for the association of APC promoter hypermethylation showing the pooled OR in cancer vs. normal controls

Full size image
Fig. 3
figure 3

Forest plot for the association of APC promoter hypermethylation showing the pooled OR in cancer vs. benign controls

Full size image

In the comparison of cancer and normal controls, subgroup analyses based on race (Asians and Caucasians) and sample type (tissue and blood) were performed to assess the difference of a strong association. According to the subgroup analysis of the ethnic population, the result suggested that APC promoter hypermethylation was significantly associated with Asians and Caucasians (OR = 8.34, 95 % CI = 3.63–19.13, P < 0.001; OR = 5.39, 95 % CI = 3.25–8.94, P < 0.001; respectively). Subgroup analysis based on sample type showed that the pooled OR value was 18.71 (95 % CI, 2.41–145.20; P = 0.005) in blood and 5.74 (95 % CI, 3.68–8.95; P < 0.001) in tissue, indicating that the result was significantly associated with different sample types.

In addition, when 319 cancer patients were compared to 75 LMP patients from 5 studies, no significant association was observed in our study (OR = 1.30, 95 % CI = 0.67–2.51, P = 0.436) (Table 2).

The association of APC promoter hypermethylation with clinicopathological features

We further analyzed the possible association between APC promoter hypermethylation and clinicopathological features. As shown in Table 2, there was statistically significant heterogeneity in APC promoter hypermethylation in ovarian cancer in relation to tumor grade and tumor stage (I2 = 68.5 %, P = 0.013; I2 = 61.0 %, P = 0.025; respectively), using the random-effects model. No obvious heterogeneity was found in relation to tumor histology, thus, a fixed-effects model was used (I2 = 0.0 % and P = 0.584).

The pooled OR from 4 studies involving 116 low-grade ovarian cancer patients and 148 high-grade ovarian cancer patients was shown in Table 2 (OR = 0.46, 95 % CI = 0.13–1.65, P = 0.233), suggesting that APC promoter hypermethylation was not significantly associated with tumor grade. As shown in Table 2, our result from 5 studies indicated that the correlation between APC promoter hypermethylation and tumor stage was not statistically significant (OR = 0.76, 95 % CI = 0.31–1.88, P = 0.558), including 188 early ovarian cancer patients and 278 advanced ovarian cancer patients. The overall OR from 7 studies including 193 serous carcinoma and 217 non-serous carcinoma cases suggested that APC promoter hypermethylation was significantly correlated with tumor histology (OR = 0.56, 95 % CI = 0.35–0.91, P = 0.02), and it was lower in serous carcinoma than in non-serous carcinoma (Fig. 4).

Fig. 4
figure 4

Forest plot for the correlation of APC promoter hypermethylation showing the pooled OR in cancer in relation to tumor histotype

Full size image

Publication bias

Egger’s linear regression test was performed to detect the potential publication bias. The results of Egger’s test on APC promoter hypermethylation indicated that there was not obvious evidence of publication bias in the comparison of cancer and control groups, in relation to tumor grade, tumor stage, and tumor histology in cancer (all P > 0.05) (Table 2).

Discussion

The loss of gene expression associated with the CpG islands of promoter methylation of different genes has been found in many cancers [33–35]. The promoter hypermethylation of tumor suppressor genes (TSG) involving cell proliferation, cell death, cell migration, and cell invasion leads to the initiation and development of cancer [36]. APC promoter hypermethylation has been reported in some cancers, including ovarian cancer, which indicates that it can become a noninvasive biomarker for cancer detection [23, 37–39]. Although some studies have been conducted to assess the frequency of hypermethylation of the APC promoter in OC, the results were still inconsistent and controversial. For example, Caceres et al. [29] found that the frequency of APC promoter hypermethylation was 11.4 % in ovarian cancer tissue. Tam et al. [27] found that frequency of APC promoter hypermethylation was 47.2 % in ovarian cancer tissue. Therefore, we utilized this meta-analysis to identify the correlation between APC promoter hypermethylation and OC.

Our findings supported that OC had a greater hypermethylation of APC promoter than benign lesions and normal samples, suggesting that APC inactivation via hypermethylation is involved in the carcinogenesis and development of OC. However, the frequency of APC promoter hypermethylation was similar in OC and LMP, indicating that APC promoter hypermethylation could not distinguish OC and LMP. Our results were credible based on the lack of publication bias.

When cancer was compared to normal controls, subgroup analysis of ethnicity showed that the pooled OR from the Asian and Caucasian populations was similar, suggesting that Asian and Caucasian populations were susceptible to APC promoter hypermethylation. Interestingly, subgroup analysis of sample type indicated that hypermethylated APC was significantly higher in blood (OR = 18.71, P = 0.005) than in tissue (OR = 5.74, P < 0.001), which suggested that APC promoter hypermethylation may become a potential noninvasive biomarker based blood test for OC. However, the results should be carefully considered as only two studies with a small number of subjects were included in the blood subgroup.

Next, the clinical significance of APC promoter hypermethylation was first determined in cancer. APC promoter hypermethylation was not correlated with clinical stage and tumor grade, but was associated with histological subtype in which it was significantly lower in serous carcinoma than in non-serous carcinoma, indicating that APC promoter hypermethylation was correlated with a decreased risk of serous carcinoma. Therefore, APC promoter hypermethylation may be a potential drug target for serous carcinoma.

The current study had several potential limitations. First, although we searched the literature as completely as possible, only articles published in English or Chinese were selected, which may lead to selection bias. Second, based on insufficient data and studies about other clinicopathological features, such as age, lymph node status, etc., the correlation between APC promoter hypermethylation and other clinicopathological features was not conducted in the present meta-analysis. Third, the interpretation of the pooled OR of the blood subgroup with a small sample size should be conservative. More studies comprising larger sample sizes are necessary to confirm our results.

Conclusion

In conclusion, the current findings revealed that APC promoter hypermethylation may play a key role in the initiation of OC. APC promoter hypermethylation decreased the risk of the serous carcinoma histotype. Moreover, hypermethylated APC may be a potential promising biomarker for the clinical screening of OC in blood. Based on the limitations of the current sample size, additional studies is very essential in the future.