Diagnostic performance of biomarkers for ovarian cancer: Protocol for an overview, evidence mapping, and adjusted indirect comparisons.

BACKGROUND: Ovarian cancer is one of the deadliest gynecological diseases and the annual mortality of ovarian cancer continues to rise. The prognosis of ovarian cancer is poor because it is prone to early metastasis during progression. Therefore, early diagnosis of ovarian cancer is very important. Some systematic reviews have evaluated the diagnostic value of different biomarkers for ovarian cancer. However, there is no consensus in the conclusions, and some are even contradictory. This study aims to assess the methodological and reporting quality of available systematic reviews and to find an optimal biomarker for diagnosing ovarian cancer.
METHODS: The PubMed, Embase.com, the Cochrane Library of Systematic Reviews, and Web of Science were searched to identify relevant systematic reviews from inception to February 2019. We included systematic reviews that include randomized controlled trials, cross-sectional studies, case-control studies, or cohort studies as long as the systematic reviews evaluated the diagnostic performance of biomarkers for ovarian cancer. The methodological quality will be assessed using assessment of multiple systematic reviews-2 checklist, and the reporting quality will be assessed using preferred reporting items for systematic reviews and meta-analysis diagnostic test accuracy (PRISMA-DTA) checklist. The pairwise meta-analysis and indirect comparisons will be performed using STATA (13.0; Stata Corporation, College Station, TX).
RESULTS: The results of this overview will be submitted to a peer-reviewed journal for publication.
CONCLUSION: This overview will provide comprehensive evidence of different biomarkers for diagnosing ovarian cancer. PROSPERO REGISTRATION NUMBER: CRD42019125880.

Introduction

Ovarian cancer is the seventh most common female cancer in the world,[ but it is one of the deadliest gynecological diseases, with 295,414 new cases and 184,799 death in 2018 worldwide.[ The incidence and annual mortality of ovarian cancer continue to rise, especially in developing countries.[ Although most patients can respond to the first treatment, the prognosis is still poor, as ovarian cancer is prone to early metastasis during its progression.[ Previous studies have shown that 5-year survival rate in patients with advanced ovarian cancer is only 30% after treatment, while 5-year survival rate in early ovarian cancer patients is as high as 92.7%.[ Therefore, there is an urgent need to identify some biomarkers for early diagnosis of ovarian cancer to improve its prognosis.

Fortunately, many scholars have devoted themselves to exploring potential biomarkers for diagnosing ovarian cancer during the past several years. The carbohydrate antigen 125 (CA125) was first evaluated in the early 1980s, but this marker has low sensitivity in the early stages of ovarian cancer.[ The study conducted by Yanaranop et al[ in 2017 indicated that the specificity of Human Epididymis Protein 4 (HE4) was 86%, and the AUC of HE4 (0.893) was higher than CA125 (0.865). Cao et al[ and Zuberi et al[ revealed that microRNAs may be a potential biomarker for the diagnosis and prognosis of ovarian cancer. However, which biomarker is the optimal option for diagnosing ovarian cancer remains unclear.

Well-conducted systematic reviews (SRs) and meta-analyses of randomized controlled trials (RCTs) are often considered the best way to obtain evidence of healthcare decisions.[ Recently, some SRs have evaluated the diagnostic value of different biomarkers for ovarian cancer.[ However, there is no consensus in the conclusions, and some are even contradictory. Thus, it is crucial to re-evaluate these SRs. The objectives of this overview are: to assess the methodological and reporting quality of available SRs; to evaluate diagnostic accuracy of biomarkers for ovarian cancer by reanalyzing the results of meta-analysis; to compare the diagnostic value of different biomarkers with adjusted indirect comparisons.

Methods

Design and registration

We will conduct an overview of SRs of diagnostic test accuracy. As a part of our project, the protocol has been registered on international prospective register of systematic review (PROSPERO) (CRD42019125880). We will follow the Preferred Reporting Items for Systematic Reviews and Meta-analysis[ statements for reporting our overview.

Eligibility criteria

Type of study

We will include SRs that include randomized controlled trials, cross-sectional studies, case-control studies, or cohort studies as long as the SRs evaluated the diagnostic performance of biomarkers for ovarian cancer. The SRs should report inclusion/exclusion criteria, adequate search strategy, sufficient details about the included studies, the diagnostic value of at least 1 biomarker, the data of diagnostic value such as sensitivity, specificity, and diagnostic odds ratio (DOR).

Type of participants

We will include ovarian cancer patients regardless of the treatment regimen and tumor staging. There are no limitations in age, race, or nationality.

Type of interventions

Any type of biomarker is used to diagnose ovarian tumor including some common tumor biomarkers and some tumor-specific biomarkers. The index test can be one biomarker or one biomarker combines with other biomarkers.

Type of outcomes

The primary outcomes are sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, diagnostic odds ratio, area under the curve, and their respective 95% confidence intervals (CIs). The second outcomes are methodological and reporting quality of included SRs, and the relative diagnostic estimates of different biomarkers.

Other criteria

SRs will be excluded from the overview including diagnostic tests of imaging modalities; SRs without meta-analysis; review protocols and methodological articles.

Search strategy

The search strategy has been developed and tested through an iterative process by an experienced medical information specialist in consultation with the review team.[ A combination of subject terms and keywords was used and make appropriate adjustments of vocabulary and grammar between different databases. The PubMed, Embase.com, the Cochrane Library of Systematic Reviews, and Web of Science were searched to identify relevant SRs from inception to February 2019. There was no restriction on the language of publication. In addition, the reference lists of included SRs have been checked for additional references. The search strategy of PubMed can be found in Supplementary 1.

Study selection

The literature search records will be imported into EndNote X8 (Thomson Reuters [Scientific] LLC Philadelphia, PA) literature management software. After removing duplicates, 2 independent reviewers will examine the title and abstract of studies found in the search to identify related studies. Then, the same 2 reviewers will retrieve the full text of all possibly relevant studies and assess the eligibility of each study according to the eligibility criteria. Conflicts will be resolved by a third reviewer.

Data extraction

To detect and resolve overlapping SRs, we will first map the research questions and characteristics of all eligible SRs. If we identify multiple reviews addressing the same research question that are eligible for inclusion but share the same primary study, we will use the following standard hierarchy to select a review to include in the overview: the review with the highest methodological quality rating; the most recent review; the review with the larger number of studies included.[ We will extract study characteristics from SRs including the following items: author name, year of publication, country of corresponding author, number of author, journal name, country of journal, funding, disease, number and name of biomarkers, number and name of reference test, and outcomes; methodological characteristics of SRs such as types of included studies, number of included studies, samples, number and name of databases retrieved, and supplemental literature search; results of statistical analysis including pooled sensitivity, specificity, likelihood ratio, predictive value, diagnostic odds ratio, area under curve, and their 95% CI. Full data abstraction will be completed by 1 reviewer and verified by a second reviewer. Disagreements will be resolved by consensus or by discussion with a third reviewer.

Quality assessment

The Assessment of Multiple Systematic Reviews (AMSTAR), published in 2007, consists of 11 items. Previous studies found that the AMSTAR is a reliable methodological quality assessment tool with good agreement, construct validity, and feasibility.[ But it was developed to evaluate SRs of randomized trials. AMSTAR-2, a major revision of the original AMSTAR instrument, could be used to assess SRs based on non-RCTs.[ Thus, we will use it to assess the methodological quality of included SRs.

The preferred reporting items for systematic reviews and meta-analysis diagnostic test accuracy (PRISMA-DTA), consists of 27 items, is an expanded checklist of original PRISMA, which aims to improve the completeness and transparency of reporting of SRs of diagnostic test accuracy studies.[ We will use it to assess the reporting quality of included SRs. Two review authors will independently assess the risk of bias in each study according to predefined criteria. Disagreements regarding by-item and overall rating of quality will be resolved by consensus or third-party adjudication if consensus cannot be reached.

Data synthesis

Evidence map

Map the biomarkers. We will create a bubble plot according to the biomarkers for all included SRs. This map displays information in 3 dimensions the bubble size represents the total number of reviews, the total number of participants included in the SRs in the x-axis, the biomarkers in the y-axis. Map the quality. The bubble plot will be produced according to the methodological and the reporting quality, where each bubble represents 1 SR. The information of 3 dimensions in the map is the bubble size represents the number of primary studies included in the SRs, the methodological quality in the x-axis, the reporting quality in the y-axis.

Pairwise meta-analysis

We will perform a pairwise meta-analysis with the data of pooled sensitivity, specificity, DOR, positive likelihood ratio, negative likelihood ratio and their 95% CI lower limit, 95% CI upper limit using Mantel-Haenszel statistical method with STATA (13.0; Stata Corporation, College Station, TX). The heterogeneity between each study will be estimated using the P value and the inconsistency index (I2 test). If the I2 is ≤50%, it suggests that there is negligible statistical heterogeneity, and the fixed effects model will be employed. If the I2 is >50%, we will explore sources of heterogeneity by subgroup analysis and meta-regression. If there is no clinical heterogeneity, the random effects model will be used to perform the meta-analysis. Otherwise, clinical heterogeneity will be explored through discussion with the review team.

Adjusted indirect comparisons

We will calculate relative diagnostic outcomes between index tests including relative sensitivity, relative specificity, and relative DOR. Then, we will conduct indirect comparisons using relative diagnostic outcomes.

Assessment of publication bias

Publication bias will be assessed per biomarker; therefore, if we have >10 SRs evaluated the same biomarker then the evidence of funnel plot asymmetry will be assessed using the Begg test using a P value of 0.1 to acknowledge the low power of this test.[

Subgroup analysis

If sufficient data are available, we will perform subgroup analysis on the basis of the age, body mass index, and ethnicity of participants; the country in which the study was conducted; the cutoff and time period of biomarkers; the quality of the SRs.

Discussion

This study will identify all relevant SRs that reported the diagnostic value of biomarkers for ovarian cancer. In addition to assessing the methodological and reporting quality of included SRs, we will also reanalyze the results of the meta-analysis using a pairwise meta-analysis and an adjusted indirect comparison. What is more, we will present the biomarkers and quality using the bubble plot, which can clearly show the biomarkers and quality of each SR. We hope this overview will find an excellent biomarker for diagnosing ovarian cancer and the results can help clinicians and patients choose an optimal diagnostic method for detecting ovarian cancer.

Author contributions

Jinyong Hua, Jing Wang, and Jiancheng Wang planed and designed the research. Jing Liu, Mengge Hua, Runjin Cai, and Muyang Li tested the feasibility of the study. Jing Wang and Jiancheng Wang provided methodological advice, polished and revised the manuscript. Jinyong Hua and Jiancheng Wang wrote the manuscript. All authors approved the final version of the manuscript.

Conceptualization: Jinyong Hua.

Data curation: Jing Liu, Mengge Hua, Muyang Li.

Funding acquisition: Jiancheng Wang.

Investigation: Jinyong Hua, Jing Liu, Mengge Hua, Runjin Cai, Muyang Li.

Methodology: Jing Wang, Jiancheng Wang.

Project administration: Jinyong Hua, Jing Wang.

Resources: Jing Liu, Runjin Cai, Jing Wang.

Supervision: Jiancheng Wang.

Validation: Jiancheng Wang.

Writing – original draft: Jinyong Hua, Jing Wang.

Writing – review & editing: Jing Wang, Jiancheng Wang.