This systematic review and meta-analysis were conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (see Supplementary Table S1) and are registered on the International Prospective Register of Systematic Reviews (registration number: CRD42024579585) (15). A comprehensive literature search was performed across four electronic English-language databases: PubMed (National Library of Medicine, Bethesda, MD), ISI Web of Science (Thomson Reuters, New York, NY), MEDLINE (via EBSCOhost), and Scopus. The following search terms, along with their corresponding MeSH terms, were used: “ovarian cancer,” “microRNA,” and “prognosis” (see Supplementary Table S2). Additionally, we examined the reference lists of relevant articles to identify further studies. Publications up to September 9, 2024, were included in the search. The starting years for the databases were as follows: PubMed (1965), Web of Science (1900), MEDLINE (1976), and Scopus (1968). The inclusion criteria were: (1) the study must be longitudinal, (2) the study must report on at least one type of miRNA, and (3) the study must provide prognostic outcomes for ovarian cancer, such as overall survival (OS) or progression-free survival (PFS). Studies were excluded if: (1) they were not published in English, (2) they were reviews, systematic reviews, conference abstracts, or case reports, or (3) hazard ratios (HRs) and 95% confidence intervals (CIs) could not be extracted or estimated from the article. As this study is a systematic review and meta-analysis based on publicly available data, and does not involve new human trials or private data, ethical committee approval was not required.

Article eligibility was assessed independently by two authors (ZH and LD), who also handled data extraction and quality evaluation of the eligible studies. Discrepancies were resolved with the help of a third person (ZW) to reach a consensus. Agreement between the authors on study selection and data extraction was measured using the agreement rate and kappa coefficient. Extracted data included the first author’s name, publication year, study design, region, type of miRNAs, test methods used for miRNA detection, number of patients, mean or median age, and study sample source. The methodological quality of the studies was evaluated using the Newcastle-Ottawa Scale (Table 1), which assesses three domains: selection (four items), comparability (one item), and outcome (three items) (16).

For the primary meta-analysis, miRNAs were categorized into “up-regulated” or “down-regulated” groups based on their reported expression levels in ovarian carcinoma compared with normal or benign tissues. This classification was essential because conducting a pooled analysis of all miRNAs together would be biologically confounded, given that they include both tumor-promoting and tumor-suppressing species with opposing functions. To ensure an accurate and consistent classification, we employed a hierarchical, evidence-based approach, as detailed below, to address the fact that not all included prognostic studies provided direct comparative data.

First, for miRNAs with well-documented roles in ovarian cancer, the expression status was adopted from a comprehensive review (2), which integrates evidence from comparisons with both normal and benign tissues. Second, when a prognostic study included in our meta-analysis directly reported a comparative analysis of miRNA expression between tumor and control tissues (normal or benign), this primary data was used to supplement and corroborate the classification established in Tier 1. For the miRNAs where both sources (the review and included prognostic studies) provided expression data, the pattern documented in the review (Tier 1) was taken as the definitive reference. In this study, no instances of contradictory expression direction were observed between these two tiers of evidence. Third, for miRNAs not covered by the aforementioned sources, targeted literature searches were conducted to ascertain their consistently reported expression profile in ovarian cancer. Their classification was then based on external evidence from independent studies.

The main outcomes of this meta-analysis were the HRs of miRNAs associated with OS and PFS in ovarian cancer. The miRNAs were categorized as either up-regulated or down-regulated based on their expression levels in ovarian cancer samples compared to normal samples (17). Additionally, the miRNAs were grouped according to their respective families. Given that many of the identified miRNAs were reported in only one or two studies, which was deemed insufficient for robust analysis, only those miRNAs that appeared in three or more publications were included in the calculation of synthesized HR of each miRNA family (12). The pooled HR for each miRNA or miRNA family was calculated by combining the log-transformed HRs and their standard errors from the included studies using the inverse variance method, which assigns greater weight to studies with more precise estimates such as those with smaller standard errors. The degree of variability due to differences between the studies included in the analysis was assessed using the I² statistic. A fixed-effects model was applied when low heterogeneity was observed (I² < 50%), while a random-effects model was used to estimate the pooled hazard ratios (HRs) and their corresponding confidence intervals (CIs) when higher levels of heterogeneity were detected. In cases of significant heterogeneity, subgroup analyses were performed to explore potential sources of variation, considering factors such as study region, sample source, International Federation of Gynecology and Obstetrics (FIGO) stage and histotype. Sensitivity analyses were also conducted to evaluate the impact of individual studies on the overall results by systematically excluding one study at a time. To assess publication bias, the Egger’s tests (18) were performed, alongside a visual inspection of funnel plots. Trim and-fill analyses were further conducted to control for potential publication biases when P-values for Egger’s tests were less than 0.05. All meta-analyses were conducted using STATA software (version 14.0) and R software (version 4.3.0). Statistical significance was set at a two-sided P value threshold of 0.05 for all analyses.

A total of 5,516 titles and abstracts were initially retrieved from the four databases. After removing 2,380 duplicate records, 3,136 records remained for further screening. Of these, 3,038 records were excluded for the following reasons: 228 were reviews, conference abstracts, comments, or letters, and 2,810 were unrelated to miRNA or ovarian cancer. After screening 98 records and conducting a full-text review, 48 studies were deemed eligible for inclusion in this systematic review and meta-analysis. Additionally, three studies were identified through reference list checks of relevant articles. As a result, 51 records were ultimately included in the meta-analysis. (Figure 1). The agreement rate between the authors for study selection and data extraction was high and moderate (κ=0.81 and 0.68).

In total, 6,916 ovarian cancer patients and 181 types of miRNAs were included in the final analysis. All included studies were longitudinal cohort studies. Among the 51 included studies, 26 originated from China (7, 19–43), 3 originated from public databases (44–46), 22 originated from other countries, including Austria (47, 48), Denmark (49), France (50), Germany (8, 51–53), India (54), Ireland (55), Italy (4, 10, 56–59), Korea (9, 60), Norway (61), Spain (62), Turkey (63) and the United States (64). Quantitative real time-polymerase chain reaction (qRT-PCR) was performed to detect miRNA expression in the majority of the included studies (48/52). Thirty-two studies detected the miRNA expression in tissue sample, 10 studies detected in formalin-fixed paraffin embedded (FFPE) sample, and the remaining 9 studies utilized blood-based samples. In terms of the type of miRNA, 12 articles reported miRNA-200 family, followed by 6 articles on miRNA-30 family and 4 articles on let-7 family. Four studies focused on ovarian cancer at FIGO stage III-IV, while five studies specifically examined the serous histotype. In terms of quality assessment, 29 articles scored 9 points, 19 articles scored 8 points, and 3 articles scored 7 points (Table 1). Thus, the quality of articles included in this meta-analysis was relatively high with a mean score 8.5.

The association between miRNAs and OS in ovarian cancer has been reported in 47 studies. Among these, 27 studies focused on up-regulated miRNAs in ovarian cancer samples compared to normal samples, while 25 studies found down-regulated miRNAs. The relationship between miRNAs and PFS in ovarian cancer was reported in 26 studies. Of these, 17 studies reported up-regulated miRNAs, and 15 studies found down-regulated miRNAs. The synthesized HR for up-regulated miRNAs associated with OS and PFS in ovarian cancer was 0.92 (95% CI: 0.82-1.02) and 0.92 (95% CI: 0.78-1.06), respectively. For down-regulated miRNAs, the synthesized HR for OS and PFS was 0.59 (95% CI: 0.50-0.69) and 0.58 (95% CI: 0.43-0.74), respectively (Supplementary Table S3).

Subgroup analyses were conducted based on region (China vs. other countries) and sample source (tissue, blood, and formalin-fixed paraffin embedded [FFPE]). Significant regional differences were observed in the association between up-regulated miRNAs and overall survival (OS) in ovarian cancer. Specifically, elevated expression of up-regulated miRNAs was significantly associated with improved OS in Chinese ovarian cancer patients compared to those from other regions (HR 0.43 [95% CI: 0.20-0.65] vs. 1.01 [95% CI: 0.90-1.11], P < 0.01). Similarly, down-regulated miRNAs were also significantly associated with better OS in Chinese patients (HR 0.35 [95% CI: 0.27-0.42] vs. 0.71 [95% CI: 0.59-0.84], P < 0.01). Regarding sample sources, a significant difference was found in the predictive effect of down-regulated miRNAs. miRNAs extracted from tissue samples showed a stronger predictive value for ovarian cancer prognosis compared to those extracted from FFPE samples (HR 0.55 [95% CI: 0.44-0.66] vs. 0.97 [95% CI: 0.65-1.29], P = 0.046). Similar results were observed for PFS in ovarian cancer. (Supplementary Table S3).

The association between miRNAs and OS in FIGO stage III-IV ovarian cancer has been explored in four studies. The serous carcinoma HR for upregulated miRNAs associated with OS and PFS in FIGO stage III-IV ovarian cancer was 0.91 (95% CI: 0.65-1.16) and 1.03 (95% CI: 0.83-1.24), respectively. Regarding histotype, five studies have investigated the relationship between miRNAs and OS in serous carcinoma. The serous carcinoma HR for upregulated miRNAs associated with OS in serous carcinoma was 0.51 (95% CI: 0.11-0.92). (Supplementary Table S4).

The possibility of publication bias of the up-regulated miRNAs in the prognosis of ovarian cancer was low, as evidenced by the Egger’s test and the symmetrical distribution of the funnel plot. In addition, potential publication biases were found in the meta-analysis of down-regulated miRNAs in the prognosis of ovarian cancer (P = 0.003 for OS and P = 0.025 for PFS). After controlling the potential publication biases by trim-and-fill analyses the results changed significantly both in OS (HR 0.82 [95%CI: 0.65-1.05]) and PFS (HR 0.92 [95%CI: 0.60-1.40]).

Among the various miRNA families, 41 have been identified (Supplementary Table S5), with 8 families selected for this meta-analysis, as their miRNAs were reported in at least three studies. These families include miRNA-181, miRNA-199, miRNA-200, miRNA-29, miRNA-30, miRNA-96, miRNA-34, and let-7 (Supplementary Table S6). Thirteen studies examined the association between the miRNA-200 family and survival outcomes in ovarian cancer. Based on previous studies, the miRNAs included in the miRNA-200 family were miRNA-141, miRNA-429, miRNA-200a, miRNA-200b, and miRNA-200c. The association between miRNA-200 family and OS was reported in 12 studies, yielding a synthesized HR of 0.66 (95% CI: 0.53-0.80). Stratified analysis by individual miRNA members showed that elevated expression of miRNA-200a (HR 0.50 [95% CI: 0.02-0.98]) and miRNA-200c (HR 0.64 [95% CI: 0.35-0.93]) were significantly associated with improved OS. Additionally, elevated expression of the miRNA-200 family was found to be significantly associated with better PFS (HR 0.68 [95% CI: 0.47-0.89]). Further stratification revealed that increased expression of miRNA-141 (HR 0.37 [95% CI: 0.15-0.58]) was also significantly linked to improved PFS in ovarian cancer (Figure 2).

Subgroup analysis showed that the elevated expression of miRNA-200 family was subsequently significantly associated with better OS in Chinese ovarian patients compared to their counterparts from other regions (HR 0.32 [95%CI: 0.06-0.58] vs. 0.77 [95%CI: 0.63-0.91], P = 0.003) (Supplementary Table S7). In addition, When we divided miRNA-200 family into 2 subgroups based on their chromosomal location, Chr1 (miRNA-200a, miRNA-200b, and miRNA-429) and Chr12 (miRNA-141 and miRNA-200c), we found that significant statistical heterogeneity still exists (Supplementary Figure S1). The results of the sensitivity analysis are presented in Supplementary Figure S2. After excluding each study, the results did not significantly change. Furthermore, low risks of publication bias were identified in synthesized HR of miRNA-200 family both in the OS and PFS of ovarian cancer, as evidenced by the symmetrical distribution of the funnel plot (Supplementary Figure S3). Moreover, the results of the Egger’s tests suggested a low probability of significant publication bias (Supplementary Table S8).

Seven studies assessed the association between miRNA-30 family and survival outcome of ovarian cancer. The association of miRNA-30 family and OS of ovarian cancer was reported by 6 studies with the combined effect size estimate (HR 0.63 [95%CI: 0.39-0.87]) (Supplementary Table S6). Stratified analysis by miRNA-30 family member types revealed that elevated expression level of miRNA-30a (HR 0.33 [95%CI: 0.16-0.50]) and miRNA-30e (HR 0.37 [95%CI: 0.04-0.70]) were subsequently significantly associated with better OS of ovarian cancer. Also, We found that elevated expression level of miRNA-30 family had a significant association with better PFS (HR 0.75 [95%CI: 0.66-0.85]). Elevated expression level of miRNA-30a (HR 0.81 [95%CI: 0.69-0.93]) miRNA-30c (HR 0.56 [95%CI: 0.26-0.87]), miRNA-30d (HR 0.57 [95%CI: 0.28-0.86]) and miRNA-30e (HR 0.66 [95%CI: 0.36-0.97]) were subsequently significantly associated with better PFS of ovarian cancer after stratified analysis by miRNA-30 family (Figure 3). The results of the sensitivity analysis are presented in Supplementary Figure S4. After excluding each study, the results did not significantly change. The Egger’s test and funnel plot showed a significant publication bias among those studies reported miRNA-30 family and OS of ovarian cancer (P = 0.001) (Supplementary Table S8). As shown in Supplementary Figure S5, after controlling the potential publication biases by trim-and-fill analyses (8 studies added), the synthesized HR changed significantly (HR 1.09 [95%CI: 0.69-1.73]).

Four studies evaluated the association between the let-7 family and ovarian cancer prognosis. The results showed that the association between the let-7 family and overall survival (OS) and progression-free survival (PFS) in ovarian cancer was not statistically significant, with hazard ratios (HR) of 0.83 (95% CI: 0.53, 1.14) for OS and 0.83 (95% CI: 0.53, 1.13) for PFS. Similarly, the combined association between miRNA-181, miRNA-199, miRNA-29, miRNA-34 and miRNA-96 family and the prognosis of ovarian cancer is also not statistically significant (Supplementary Table S6).