Four reviewers divided into two groups and then searched four databases independently for articles related to OC and HIC1 methylation. The two groups combined their results after assessment and discussion. Original articles published up to 2020 were identified by the use of search terms including ‘ovarian’ AND ‘cancer OR tumor’ AND ‘HIC1’ AND ‘methylation’. The search process was performed without restrictions on the publication year or language.

The title, abstract and keywords from 591 studies originally identified by the search (130 from PubMed, 188 from EMBASE, 63 from Medline and 210 from CNKI) were further scanned to identify studies on OC and HIC1 methylation and to exclude irrelevant studies and reviews. To prevent inclusion bias, studies from the primary screen were downloaded for full-text review. Eligible studies were required to meet the following inclusion criteria: (A) OC patients were diagnosed by pathology; (B) studies evaluated the association between OC and HIC1 methylation; (C) the number of OC cases in the study was $>$20. Studies were eliminated if they met one or more of the following exclusion criteria: (A) reviews; (B) cases only included cell lines and animals; (C) cases without controls. After the removal of duplicates from the databases, the remaining studies were selected for data extraction and quality assessment.

Information extracted from the eligible studies was as follows: first author, year of publication, country and region of study, study population, sample size, frequency of HIC1 methylation in case and control groups, cancer stage, methylation detection method, and the tissue type used for controls. The Newcastle-Ottawa Scale (NOS) was used to assess the quality of eligible studies. A study that met one of the 9 items listed above scored one point, with the highest possible score being 9. If a study did not include one of the 9 items, that point was considered to be lost. Only studies that reached a score of 5 points or more were included in the meta-analysis.

Statistical variables were calculated using STATA (Version.12.0, StataCorp LLC, TX, USA). Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated to assess the strength of association between HIC1 methylation and OC. Methylation profiles (Illumina Human Methylation 27) and the corresponding clinical data set for 11 OC cases and 11 controls were downloaded from The Cancer Genome Atlas (TCGA) (https://portal.gdc.cancer.gov/). Heterogeneity was evaluated with Chi-square test tests and the I${}^2$ value. I${}^2$ = 0–50% indicates no or moderate heterogeneity, while I${}^2$ $>$50% indicates significant heterogeneity. The fixed-effect model was used if there was no significant heterogeneity, otherwise the random effects model model was used. Subgroup analysis was performed to evaluate the source of heterogeneity. Begg’s test and Egger’s test were performed to assess the possibility of publication bias. Asymmetry of Begg’s funnel plot and a p-value in Begg’s test of 0.05 were considered to indicate the existence of publication bias.

Following the search strategy described above, 591 studies were initially identified from the databases. Two separate review groups independently identified the studies for exclusion and then combined their results. A total of 174 studies were excluded due to being duplicates. After scanning the title and abstract of the remaining 417 studies, 398 were deemed not applicable to this research (237 irrelevant articles, 154 reviews, 4 meta-analyses and 3 others). Next, the full text for the 19 potentially relevant articles was scrutinized. Of these, 6 articles contained no data on HIC1, 3 were not focused on HIC1 methylation, one was not a full text article, one was not written in English and another did not include controls. Following exclusion of these 12 papers, 7 papers were deemed eligible for final quantitative assessment (Fig. 1).

Fig. 1.

Flow chart of study selection.

Relevant information for the 7 studies is presented in Table 1 (Ref. [9, 11, 16, 17, 18, 19, 20]). All were published between 2001 and 2013 and included a total of 733 samples (455 cases and 278 controls). The sample size in each study ranged from 46 to 207 and the majority of patients were aged between 30 and 60 years. Two studies did not record patient age and one study [16] did not report the cancer stage. In the 6 studies that reported cancer stage, there were more early stage patients than advanced stage patients. Cases were either fresh cancer tissues obtained surgically from cancer patients, or archival cancer specimens. All studies used methylation-specific polymerase chain reaction (MSP) for the detection of HIC1 methylation. The control group was comprised of adjacent tissues (AT) from OC patients, ovarian tissues from benign OC patients (BT), borderline tissues from OC patients (BLT), normal ovarian tissues (NT) and neoplasia of low malignant potential (LMP).

No significant heterogeneity was found using the Chi-squared and I${}^2$ tests (p = 0.456, I${}^2$ 0.000). A fixed-effects model was used to analyze the association between HIC1 methylation and OC. The forest plot showed the frequency of HIC1 methylation was strongly associated with OC (OR = 4.306, 95% CI = 2.846 to 6.515) (Fig. 2).

Fig. 2.

The frequency of HIC1 methylation was associated with OC.

Subgroup analysis was performed according to the control type. The ORs for the association between HIC1 methylation and OC were: AT, OR = 2.514; BT, OR = 5.944; BLT, OR = 1.375; NT, OR = 4.931; LMP, OR = 2.061 (Fig. 3). The pooled OR was 4.143 (95% CI: 2.861 to 5.999). The heterogeneity detected for control type was acceptable (p = 0.147, I${}^2$ = 41.1%).

Fig. 3.

The ORs for the association between HIC1123 methylation and OC.

Publication bias was assessed using Begg’s test and Egger’s test. As shown in Fig. 4, no significant publication bias was found, with the shape of the Begg’s funnel plot being approximately symmetrical. The p-values from the Egger’s test showed no publication bias in any comparison.

Fig. 4.

Analysis for Publication bias.

To further explore the relationship between HIC1 methylation and OC, we evaluated publicly available methylation data for OC and adjacent tissues. As shown in Fig. 5, HIC1 methylation was more frequent in OC tissues than in normal controls (p 0.01).

Fig. 5.

TCGA data showing HIC1 methylation in OC tissues and in normal adjacent tissues. Data were shown as mean $\pm$ SD. **p 0.05.