Early Ovarian Cancer Detection in the Age of Fallopian Tube Precursors: A Systematic Review

Precis:

Most biomarkers being tested for the detection of fallopian-tube-derived ovarian cancer, when combined with CA125, have similar or improved detection rates in early-stage disease.

Objective:

To determine biomarkers other than CA125 that could be used in identifying early-stage ovarian cancer.

Data Sources:

Ovid MEDLINE ALL, EMBASE, Web of Science Core Collection, ScienceDirect, Clinicaltrials.gov, and CAB Direct were searched for English language studies between January 2008 and April 2023 for the concepts of high grade serous ovarian cancer, testing, and prevention or early diagnosis.

Methods of Study Selection:

Identification of 5523 related articles were uploaded to COVIDENCE. Screening by two independent reviewers of the article abstracts led to the identification of 245 peer-reviewed primary research articles for full-text review. Following full text review by those reviewers led to the identification of 131 peer-reviewed primary research articles that were used for this review.

Tabulation, Integration, and Results:

131 studies, only 55 reported either sensitivity, specificity, or area under the curve (AUC), with 36 of the studies reporting at least one biomarker with a specificity of ≥ 80% specificity and/or ≥ 0.9 AUC.

Conclusion:

These findings suggest that while many types of biomarkers are being tested in ovarian cancer, most have similar or worse detection rates when compared to CA125 and have the same limitations of poor detection rates in early-stage disease. However, 27.4% of articles (36 of 131) reported biomarkers with better sensitivity and AUC >0.9 when compared to CA125 alone and deserve further exploration.

INTRODUCTION

Ovarian cancer is rare with a 1.1% lifetime risk however, it is the leading cause of death among gynecologic malignancies in the United States. High grade serous carcinoma (HGSC) is the most common histological subtype. It is generally accepted that the origin for HGSC is the fallopian tube epithelium (FTE). While there were reports dating back to 2001 describing the FTE as the primary site of origin for HGSC, it not was until 2007 that the understanding of HGSC’s origin shifted as detailed in two reviews (1)(2). More than 75% of patients are diagnosed with stage III/IV disease. Most patients experience no or non-specific symptoms (3, 4). The frequent late-stage of diagnosis is a consequence of limited markers for detection at early stages (I/II) (3, 4).

Currently, there are no clinical tests that can reliably detect early-stage ovarian cancer. The data from the United Kingdom Collaborative Trial of Ovarian Cancer Screening (UKCTOCS) study showed that strategies such as CA125 testing and ultrasound in the general population stage shifted ovarian cancer diagnosis to earlier stages but did not improve overall survival (OS), highlighting the need to identify biomarkers or modalities for detecting precursor lesions such as the serous tubal intraepithelial carcinoma (STIC) associated with HGSC (5). Secondly, once an ovarian mass is detected, there are significant limitations in the ability to preoperatively risk stratify ovarian neoplasms as this would limit “low risk” patients to unnecessary surgical procedures (6, 7).

Currently, serum tests such as Cancer Antigen 125 (CA125/MUC16) and Human Epididymis Protein 4 (HE4) are used to assist with preoperative ovarian neoplasm risk stratification. CA125 is a glycoprotein and HE4 is a broad protease inhibitor both of which are detected in serum of patients with ovarian cancer, however; neither CA125 nor HE4 alone have adequate sensitivity or specificity to detect early stages of the disease (8). The American College of Obstetricians and Gynecologists (ACOG) recommends referral to a gynecologic oncologist in patients with adnexal masses and a CA125 of 35 U/ml in postmenopausal patients and 200 U/ml in premenopausal patients (9). The sensitivity and specificity of CA125 are better in post-menopausal individuals at 88.7% and 98.1% compared to 64.0% and 94.1% in premenopausal patients (10). Confounding ovarian cancer detection, CA125 may be elevated for reasons unrelated to malignancy, such as endometriosis, menstruation, physiologic states, pregnancy, or anything that would irritate the peritoneal lining, thus decreasing its sensitivity (8).

HE4 is commonly overexpressed in epithelial-derived ovarian tumors. In a meta-analysis by Wang et al., HE4 and CA125 had similar abilities to discriminate malignancy from benign mass with Area Under the Curves (AUC) of 0.89 for HE4 and 0.87 for CA125. HE4 had a higher specificity than CA125, especially in the premenopausal subgroup, 93.8 vs 76.3, respectively, whereas CA125 performed better in the post-menopausal group when compared to HE4 (11). While HE4 shows potential improvement in premenopausal patients as a single marker, the overall similar detection rate in the general population, high cost of the test, and limited availability in certain areas limits its use (12).

Given the limitations of any single marker, both HE4 and CA125 have been evaluated in combination to evaluate risk of malignancy. For example, HE4 and CA125 are used in the Risk of Ovarian Malignancy Algorithm (ROMA) (13), which is a numerical score used to predict risk of epithelial ovarian cancer in patients with an adnexal mass. In a meta-analysis, ROMA performed similarly to CA125 with sensitivity between 76 and 86%, while specificity was between 74 and 95%. The AUC for the ROMA algorithm was better than HE4 or CA125 alone at 0.93, versus 0.82 and 0.88, respectively (14). The Risk of Malignancy Index (RMI) is another risk assessment tool that is referenced which utilizes menopausal status, ultrasound findings, and CA125 levels. A RMI score above 200 proved to be a good predictive model for classifying a patient with an adnexal mass as high risk for malignancy with a sensitivity of 87.5%, specificity of 90.7% and AUC of 0.9 (15).

With the limitations in the current biomarker testing the need for a better marker remains a challenge. The goal of this systematic review is to explore the recent literature for promising tests that could aid in the detection of ovarian cancer, particularly in the setting of early-stage disease and precursor lesions, for which there is a paucity of effective testing.

SOURCES

Search terms and criteria are detailed for each of the five repositories queried in the supplemental information section (Appendix1, available online at http://links.lww.com/xxx). Five repositories were queried to complete a comprehensive search and to avoid missing relevant articles. Clinicaltrials.gov was independently queried. Eligibility criteria of literature were determined a priori. Included studies had to include information about high grade serous ovarian cancer and meet one of the following criteria: Identifies a biomarker/method that correlates/associates with diagnosis of disease, identifies a biomarker/method that correlates/associates with early-stage disease, identifies a biomarker/method that correlates/associates with the diagnosis of early-stage disease, identifies a biomarker/method that correlates/associates with disease progression, or identifies a biomarker/method that correlates/associates with the transformation of fallopian tube epithelium. Exclusion criteria of publications included prior systematic reviews or reviews of the literature reporting biomarkers or detection methods, articles in languages other than English, articles published before 2008, articles that discuss HGSC but identify a biomarker or method that correlates or associates solely to therapy response, neoadjuvant response, therapy resistance, or prognostication.

A comprehensive literature search was designed and performed by a medical librarian (CP) in January 2022 for the concepts of high grade serous ovarian cancer, testing, and prevention or early diagnosis. Relevant publications were identified by searching the following databases with a combination of standardized index terms, when available, and keywords: Ovid MEDLINE ALL (1946 to January 5, 2022), Embase (via Elsevier, 1947 to present), Web of Science Core Collection (via Thomson Reuters, including Science Citation Index Expanded 1974 to present, and Social Sciences Citation Index 1974 to present), ScienceDirect (Elsevier) Journals & Books, and CAB Direct (including CAB Abstracts and Global Health; Last updated on January 4, 2022). Searches were developed in Ovid MEDLINE and translated to additional databases. Results were limited to publication dates from 2008 to present and English language, and systematic reviews and reviews were excluded when possible. Duplicates were removed using Covidence systematic review software (Veritas Health Innovation), which was also used for screening and full text review. See Supplemental Materials for a complete list of all database search strategies. The search was rerun in April 2023 to identify any new publications and the concept of “serous” was removed from the search strategy. The systematic review has not been registered in PROSPERO. All included articles were included in the updated search, so only the updated search strategy is accounted for in the PRISMA diagram and Supplemental Materials.

Citations and abstracts were uploaded in Covidence for study selection. As measure of certainty or confidence, all the selected articles were confirmed to be peer-reviewed. Two authors (BGB, ERW) independently screened all titles and abstracts. Articles considered for inclusion were independently reviewed by two authors, and consensus was reached by discussion on any conflicting articles selected for inclusion. While both reviewers are ovarian cancer specialists, there is noted potential for selection bias of the reviewers due to an emphasis in their basic science training.

Literature was compiled in Covidence (Veritas Health Innovation Ltd, Melbourne, AUS), PRISMA figure was generated using Covidence, and tables were developed using Microsoft Excel.

Sensitivity is the ability of a positive test to correctly identify an individual with the disease being tested. Specificity is the ability of a negative test to correctly identify an individual who does not have the disease being tested. The area under the curve (AUC) is taken from a receiver operating characteristic curve (ROC curve). It is a way to quantify the ability of a test to determine a diseased vs non-diseased person. An AUC of 1.0 would be able to distinguish diseased vs non-disease perfectly whereas AUC 0.5 would be no better than chance in determining disease vs non-disease (16).

RESULTS

In the five major repositories, 5523 related articles were uploaded to COVIDENCE. After the removal of duplicates, 3747 related articles matched our search terms (Appendix 2, available online at http://links.lww.com/xxx). Further screening by two independent reviewers of the article abstracts led to the identification of 245 peer-reviewed primary research articles for full-text review. Following full text review by two independent reviewers and removal of duplicate articles led to the identification of 131 peer-reviewed primary research articles (Figure 1 PRISMA Flow diagram, Table 1 and Appendix 3, available online at http://links.lww.com/xxx). Note, there were seven clinical trials identified via the ClinicalTrials.gov query and three were directly related to early diagnosis (NCT04794322, NCT05146505, and NCT03622385), however none had reported outcomes. Given the expertise of the reviewers, the systematic approach for article selection, and the a priori criteria, the 131 articles identified in the five literature repositories are deemed to be high confidence selections. Thus, these 131 articles serve as the source of the remainder of this systematic literature review.

Figure 1:

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram for new systematic reviews which included searches of databases and registers only. *All records excluded were due to selection by two independent reviewers. Any conflicts were discussed, and final decision was made by one reviewer (BB).

TABLE 1.

Study Design (n=131)
Retrospective	120
Prospective	9
Both	2
Comparison
Healthy versus Cancer	37
Precancer versus Cancer	12
FTE versus Cancer	9
OSE versus Cancer	7
Multicomparisons	43
BRCA Wildtype versus BRCA1 Mutated	1
Healthy versus Precancer	2
Mixed Comparisons	20
Number of Specimens
Normal/Precancer; Sum of n (Min., Max)	98496 (3, 40941)
Cancer (n); Sum of n (Min., Max)	64038 (3, 25509)
Factor Examined/Biological Source
Tissue	43
Blood	46
Urine	2
Proximal Fluid	5
Medical Record	1
Imaging	0
Publicly Available Data (e.g., TCGA)	1
Cell Line	1
Meta-Data Analysis	1
Mixed	31
Readout
Clinical Attribute	2
Protein	42
RNA	23
DNA	11
Metabolite	4
Post-translational Modification	1
Chromatin	0
Cytology	1
DNA Methylation	4
Fluorescence	1
Iron/Zinc	1
Vibrational Spectral Absorbance	3
Mixed (2 Readouts)	32
Mixed (3/4 Readouts)	6

As of 2007, early-stage high grade serous carcinoma was appreciated to predominantly occur in the fallopian tube. Consistently, a review of the 131 articles revealed that over 88.5% of the articles that examined or used FTE as a comparator to identify novel strategies for HGSC early detection were published after 2015 (Appendix 4, available online at http://links.lww.com/xxx). The 131 articles were published in 78 different peer-reviewed journals, with Gynecologic Oncology representing the majority at 7.6% (10 out of 131).

The design and results of the 131 selected articles were further examined. A summary of the study selection and characteristics can be found in Table 1. Nearly 91.6% (120 of 131) of the articles were retrospective and examined specimens already collected, while 6.8% (9 of 131) were prospective studies (studying characteristics that predate diagnosis) to evaluate an early-stage biomarker. Note, two studies (1.5%) conducted both retrospective and prospective studies. The sample size across the 131 studies ranged from 6 (17) to 66,450 (18) Most studies only reported a discovery cohort, while 34 studies validated their findings either through independent datasets (20 studies) or cross-validation methods (14 studies). Of the 131 studies, 55 reported either sensitivity, specificity, or area under the ROC curve (AUC), with 30 of the studies reporting at least one biomarker with an AUC >0.9. These findings suggest that appropriate methods for biomarker identification and classification remain underutilized.

Over 45.8% (60 of 131) of the studies examined a protein, peptide, or post-translation protein modification as a biomarker for early detection, and some studies aimed specifically to identify new markers that when used in combination would improve the diagnostic performance of CA125 and/or other conventional tests. There are 24 studies that measured only a single protein and none of these studies reported a sensitivity of greater than 80%. Urunsak et al. examined serum adenosine deaminase (ADA) and was able to differentiate between ovarian cancer versus benign tumors at 84% sensitivity and 80% specificity (AUC 0.82); however, peritoneal fluid ADA did not perform as well (19). Another study examined plasma-derived Annexin A2 (ANXA2) and reported a sensitivity of 80% at 99.6% specificity for distinguishing HGSC Stage IA vs healthy controls when combined with CA125 (AUC 0.970 when combined with CA125, AUC 0.774 = ANXA2 alone).

There are 30 identified studies that examined multiple proteins alone or with other biomarkers including metabolites and miRNAs. Combination of multiple biomarkers tended to improve sensitivity or AUC, as 12 of 30 studies reported >80% sensitivity and/or AUC of greater than 0.9 (Table 2). Huh et al. assessed an 18-protein model that showed sensitivity of 100% and specificity of 91% in distinguishing HGSC vs healthy patients (AUC 0.99) (20). Kampan et al. assessed serum levels of IL-6 and HE4 in patients with either HGSC ovarian cancer, a benign mass, or normal ovaries and reported sensitivity and specificity at 100% with an AUC of 1.0 (12). Five studies examined post-translation protein modification where only one reported AUC (21) that reached a level greater than 0.9 when a panel of anti-proteoglycans (5 glycans) was combined with CA125 testing.

TABLE 2. Published studies with a reported specificity (≥ 80%) or Area Under the Curve (AUC, ≥0.9).

Primary readout focus	Biospecimen	Effect on marker(s)	AUC (95% CI)	Sensitivity (%, 95% CI)	PMID
Protein, Peptides, or Post-translational Modification	Tissue	Tissue TOP1, PDIA4, and OGN expression profiles highly discriminatory		98.2	35197484
Protein, Peptides, or Post-translational Modification	Blood	Panel of CA125, HE4, E-CAD, and IL-6 distinguished early stage HGSC from non-malignant controls with higher efficacy than CA125, HE4, or CA125+HE4	0.961 ± 0.0243	84.2	29572027
Protein, Peptides, or Post-translational Modification	Blood	84 up-regulated, 32 down-regulated proteins in serum from HGSC patients vs. healthy controls.	0.99	100.0	35939567
Protein, Peptides, or Post-translational Modification	Blood	Serum IL-6 distinguish between HGSC, benign ovarian masses, and non-malignant controls. Diagnostic value highest when IL-6 combined with CA125 and HE4.	IL-6: 0.962 (0.926–0.998) IL-6 + CA-125: 0.985 (0.966–1.000) IL-6 + HE4: 1.000 (1.000–1.000)		32042020
Protein, Peptides, or Post-translational Modification	Blood	Combination of TNRF2+ Tregs and IL-6 in blood of advanced stage HGSC discriminates benign ovarian masses and non-malignant controls.	1.000 (1.000–1.000)		36765633
Protein, Peptides, or Post-translational Modification	Blood	Machine learning was able to predict EOC diagnosis and EOC stage based on several blood-born CRP, lymphocyte count, and CA125.	Conditional random forest: 0.978 Gradient based machine: 0.976 Random forest: 0.968		30979733
Protein, Peptides, or Post-translational Modification	Blood	Plasma ANXA2 elevated in early stage (I and II) HGSC compared to non-malignant controls. ANXA2 with CA125 test highly diagnostic of early stage OC.	0.969	84.4	33406648
Protein, Peptides, or Post-translational Modification	Blood	Serum Protein Z, fibronectin, CRP, and CA125 effective in predicting OC occurrence 2–3 years pre-diagnosis.	0.944 (0.896–0.992)		27903971
Protein, Peptides, or Post-translational Modification	Blood	Incorporating longitudinal measurement of serum CA125, HE4, CHI3L1, PEBP4, and/or AGR2 predicted OC (particularly HGSC) up to one year before diagnosis.	CA-125 + PEBP4: 0.97 (0.934–1.000) CA-125 + CHI3L1: 0.986 (0.973–0.999) CA-125 + AGR2 + CHI3L1: 0.984 (0.971–0.998) CA-125 + HE4: 0.988 (0.976–1.000)	CA-125 + PEBP4: 95.5 (77.3–100.0) CA-125 + CHI3L1: 100.0 (90.9–100.0) CA-125 + AGR2 + CHI3L1: 100.0 (90.9–100.0) CA-125 + HE4: 100.0 (86.4–100.0)	31937926
Protein, Peptides, or Post-translational Modification	Blood	Plasma antibodies against HSF1 and CCDC155 are elevated in early stage HGSC and are superior to CA125. Combined measurement improved sensitivity and efficacy of detection.	HSF1: 0.95 CCDC155: 0.80		29141850
Protein, Peptides, or Post-translational Modification	Blood	Increased serum MMP-9, Hpa, and CL levels in OC vs. healthy control and benign ovarian mass patients; low grade and advanced stage vs. high grade and early stage patients.	0.935	96.4	23359763
Protein, Peptides, or Post-translational Modification	Proximal Fluid	Increased GJA1, C4BPB, ATP2B4, VPS11, and TMEM67 expression and decreased KIF20B expression in HGSC proximal fluid vs. non-malignant control proximal fluid.	C4BPB + KIF20B: 0.979 (0.953–1.00) VPS11 + CRTAC1 + TMEM67: 0.968 (0.938–0.999) GJA1 + ATP2B4: 0.943 (0.889–0.997)		36214786
Protein, Peptides, or Post-translational Modification	Blood, Proximal Fluid	Elevated serum and ovarian cyst fluid ALDOA diagnostic in early stage EOC (low grade and high grade) with (LC)-MS/MS.	0.96		30710757
Protein, Peptides, or Post-translational Modification	Blood, Proximal Fluid	Serum and proximal fluid ADA upregulated in HGSC patients vs. patients with benign ovarian mass.	Serum ADA: 0.82 Peritoneal fluid ADA: 0.78	Serum ADA: 84.0 Peritoneal fluid ADA: 74.0	22395862
Protein, Peptides, or Post-translational Modification /DNA Mutational Profiles	Blood	Increased circulating histone-DNA complex, cfDNA, neutrophil elastase, prekallikrein, and CA125 in HGSC patients vs. healthy controls.	0.966 (0.933–1.000)	97.3	36620601
Protein, Peptides, or Post-translational Modification /DNA Mutational Profiles	Blood, Proximal Fluid	Increased NETosis biomarkers (cfDNA, nucleosomes, cirtullinated histone 3, calprotectin, and myeloperoxidase) in serum and peritoneal fluid from HGSC patients vs. healthy controls.	cfDNA: 0.90 (0.80–1.00) Nucleosomes: 0.94 (0.87–1.00) citH3: 0.96 (0.92–1.00) Calprotectin: 0.91 (0.84–1.00) MPO: 0.87 (0.77–0.98)		36817483
Epigenetics	Tissue	Whole methylome sequencing identified novel OC methylated-DNA- markers; differentiated 63/73 HGSC cases included 5/5 stage I/II cases.	0.91 (0.86–0.96)	79 (69–87)	35370009
Epigenetics	Tissue	Elevated methylation in promoter regions of TUBB6, IRX2, and c17orf64 in HGSC and precursor STICs compared to non-malignant controls.	1.0	100.0	30108103
Epigenetics	Tissue	Methylation landscape of STICs intermediate between normal FTE and HGSC tumors.	PCDHB12: 0.958		32817081
Epigenetics	Blood	Serum miR200a, b, and c higher in patients with serous EOC vs. non-malignant controls. Combo miR200b + c best predictive qualifier.	miR-200a: 0.675 miR-200b: 0.722 miR-200b + c: 0.784	miR-200a: 85.7 miR-200b: 85.7 miR-200b + c: 78.6	23272653
Epigenetics	Blood	Upregulated serum exosomal miR-93, miR-145, and miR-200c in HGSC samples compared to non-HGSC cases, benign, and borderline groups. Specificity and sensitivity superior to CA125.	miR-145: 0.910 (0.840–0.980) miR-200c: 0.802 (0.698–0.906)	miR-145: 91.7 miR-200c: 72.9	31205555
Epigenetics	Blood	Model with 18 differentially methylated DNA regions (cfDNA) able to differentiate OC from non-malignant patients.	0.967 (0.940–0.994)	94.7 (85.4–98.9)	35973389
Epigenetics	Blood	miR-1246 overexpressed in both serum and tumors of HGSC patients vs. non-malignant controls.	0.893	87	28017893
Epigenetics	Blood, Cell line	Serum miR1290a elevated in HGSC compared to other histotypes and non-malignant controls. Serum levels more effective at detection than CA125 alone and positively associated with FIGO stage.	miR-1290: 0.71 CA-125 + miR-1290: 0.97	miR-1290: 63	30219071
Epigenetics/RNA	Tissue	Lower CDH13, HNF1B, PCDH17, and GATA4 gene expression in HGSC tumors vs. non-malignant control tissue, particularly in those samples with high gene methylation.		88.5	32145055
Epigenetics/RNA	Blood	RASSF1A promoter methylation increased in EOC vs. healthy control; HGSC vs. LGHOC, advanced stage vs. early stage.	Serum RASSF1A: 0.993 (0.96–0.99)	Serum RASSF1A: 97 RASSF1A promoter methylation: 85	29098560
DNA Mutational Profiles	Tissue	Using 49 different Copy Number Variant loci, can differentiate EOC from FT, AUC 1.0.	1.00 (1.00–1.00)		36499142
DNA Mutational Profiles	Blood	Comparing HGSC blood Copy Number Index-Score to that of healthy controls detects cancer.		91.0	35008332
DNA Mutational Profiles	Blood	Higher plasma cfDNA in OC vs. non-malignant controls, positively associated with copy number alterations and FIGO stage.	0.94	78.0	27852697
DNA Mutational Profiles	Blood	Alterations in circulating cfDNA can be combined with CA125 to improve differential diganosis of OC.	0.9752	96.0	34053311
Metabolite	Blood	Alterations in several mouse serum metabolites detectable, differentiating between non-HGSC samples (control mice), early, and advanced HGSC (triple KO mice)		96.2	31290664
Metabolite	Blood	Serum levels of 147 lipid species between HGSC patients and healthy controls; 100 species different between Stage I/II and controls. Lipid levels also varied by disease stage (I/II vs. III/IV).	1.0	100.0	35495636
Metabolite	Urine	Increased urine N1,N12-diacetylspermine levels in OC vs. tumor-free controls, HGSC vs. low malignant. Potential patients, stage III/IV vs. stage I/II, stage I/II vs. benign tumor patients.	0.83	86.5	28604456
Metabolite	Blood, Tissue	Increased plasma C16-Cer, C18:1-Cer, C18-Cer in HGSC ((+):: FIGO stage); increased tissue C16-Cer, C18:1-Cer, C18:Cer, C24:1-Cer, C24-Cer & SIP and decreased SPH in HGSC vs. normal tissue.	C18:1-Cer: 0.768 (0.53–0.81) C18-Cer: 0.771 (0.53–0.8) C16-Cer: 0.759 (0.51–0.8)	C18:1-Cer: 90.0 (59.6–98.5) C18-Cer: 80.0 (50.0–94.7) C16-Cer: 77.0 (56.5–94.3)	28800942
Metabolite/RNA	Blood, Tissue	Increased circulating lactate in HGSC patients, increased HCAR1 mRNA and protein expression in ovarian cancer tissue vs. healthy control.	0.969 (0.940–0.998)		36615018
Metabolite/RNA	Blood, Tissue, Public Dataset	Increased hydroxybutyric acid metabolites in sera and tumors from HGSC patients compared to non-malignant controls.	0.91		26685161

Abbreviations: CA125 - cancer atnigen 125 or MUC16, EOC - epithelial ovarian cancer, FIGO - International Federation of Gynecology and Obstetrics, FT - fallopian tube, FTE - FT epithelium, HGSC - high grade serous carcinoma, mRNA - messenger RNA, miR - microRNA, OC - ovarian cancer

Epigenetic changes are defined by a modification or regulation of genetic programming without changes to the underlying genetic sequence. Numerous studies have demonstrated aberrant epigenetic regulation in the process of cellular transformation to cancer. In our literature review, 27 studies incorporated an epigenetic biomarker to aid in the detection of early-stage HGSC. Eleven studies examined DNA methylation, one study examined transfer Ribonucleic Acid (RNA), and one study examined PAX8 (Mullerïan marker)-DNA binding. Pisanic et al, reported HGSC-specific differentially-hypermethylated regions with AUC>0.9, and specifically hypermethylation on the PCDHB12 gene reported an AUC 0.958 when STIC vs paired adjacent-normal FTE and C17orf64 gene an AUC 0.924 HGSC vs healthy FTE (22). Pisanic et al, observed hypermethylation of two genes loci identified with AUC>0.9, c17orf64 (AUC 0.968) and IRX2 (AUC 0.928) for detecting HGSC compared to samples from healthy individuals. Moreover, the combination of c17orf64 + IRX2 + TUBB6 (top performing three markers) resulted in an AUC of 1.0 for detecting HGSC compared to samples from healthy individuals (23).

MicroRNA (miRNA, miR) accounted for 14 of these studies, six of which examined members of the miR-200 family (e.g., miR-200c). Four studies reported sensitivity ≥80% and/or an AUC ≥0.9 for detecting cancer vs non-cancer (24, 25) and only one reported an AUC ≥0.9 (24). Note, an on-going clinical trial, NCT05146505, is leveraging miRNA as an early diagnostic tool.

During the progression of FTE to HGSC, there are appreciable genetic mutations that are predicted to occur in nearly all HGSC cells. For instance, mutation of TP53 (p53) tumor suppressor is nearly ubiquitous in HGSC tumors, with 75 to 98% of HGSC tumors harboring a p53 mutation (26). Also, nearly 50% of HGSC tumors harbor mutations in genes involved in DNA double strand break (DSB) repair (e.g., BRCA1, BRCA2, PALB2) (26, 27). The consequence of impaired DNA DSB repair is high frequency of chromosomal instability (e.g., amplifications, translocations, etc.). In the literature we reviewed, 14 studies examined gene mutations with six studies reporting sensitivity ≥ 80% and/or AUC ≥ 0.9. In Erickson et al., p53 mutations were determined in blood isolated from tampons of patients with or without HGSC, and while all 8 HGSC tumors harbored p53 mutation in the tumor, only 3 of the 8 blood samples from tampons showed the mutation (28). In Gonzalez-Bosquet et al, a model of 49 single nucleotide variants had excellent performance with AUC of 1.0 in distinguishing HGSC from benign fallopian tube; models with 11 copy number variants (AUC 0.87) and 17 structural variants (AUC 0.73) performed more poorly (29). In Vanderstichele et al., chromosomal copy number from cell free DNA was examined in a total of 112 patients (44 healthy control, 57 HGSC or borderline, 11 benign) and they reported a specificity of 99.6% and sensitivity of 78% when comparing benign tumors to HGSC tumors (30). As sequencing technologies continue to become more cost-effective with increased sensitivity the ability to scale up sequencing-based biomarkers becomes more attainable.

In addition to DNA, RNA and RNA profiles have been proposed to be viable biomarker for cancer progression. Thirteen studies evaluated multiple gene expression, and two studies examined circular RNA expression; however, only two studies reported sensitivity/specificity and AUC, and neither of these reported sensitivity ≥ 80% or AUC ≥ 0.9. Notably, Dinh et al. performed single cell RNA sequencing on 12 normal fallopian tubes and identified 10 distinct sub-populations for epithelial cells. The investigators further used the underlying transcriptomic profiles to develop a differentiation trajectory between the different cell types. Next, using the transcript profiles for each population, the investigators were able to deconvolute RNA-sequencing data from HGSC tumors and to identify the precursor epithelial sub-population responsible for the HGSC tumor. This highlights the use of both sequencing and advanced computational analysis to further elucidate the transformation of FTE to advanced HGSC. Furthermore, this study highlights the power of sequencing nucleic acids from the serum or tumor compartment for diagnostic purposes and thus remains a major area of research.

Cancer-associated metabolic reprogramming offers a unique opportunity to track disease progression and serve as a biomarker for early-stage disease. Notably, reports have shown an increased dependence on lipid metabolism during HGSC progression (31–33). In the literature reviewed here, seven studies evaluated metabolites in serum, tissue, and/or ascites, and all but one reported sensitivity and/or AUC, all reported sensitivity and AUC ≥ 80% and/or ≥ 0.9, respectively. For instance, Niemi et al. performed lipidomic analysis on the sera of 354 patients (Malignant n=138, Borderline n=25, Healthy Control n=191). The investigators observed 39 lipid species elevated in both early- and late-stage disease. Lipid species increased with increasing stage (34). Ceramide (d18:1/18:0), a type of sphingolipid, was notably elevated in both HGSC tumors and pre- versus post-menopausal individuals. The rise in lipids seemed to be restricted to the HGSC histotype. Further, in stage I/II disease combining lipid profiles with CA125 reported an AUC of 0.87 compared to CA125 alone that reported an AUC of 0.69. As discussed below, the ability to serial test through a non-invasive blood draw is an attractive approach for the development of an early HGSC diagnostic tool.

DISCUSSION

The ability to detect early-stage ovarian cancer has represented a significant area of research for the last four decades. In parallel to researching suitable biomarkers, the understanding of ovarian cancer etiology, tumorigenesis, and progression has significantly advanced, especially in the last 15 years. As the fallopian tube is appreciated to be the primary site of HGSC tumorigenesis, there has been a shift in the early-stage biomarker research to include normal fallopian tube or adjacent normal fallopian tube as a comparator against HGSC tumors. In this systematic review, there was an attempt to summarize the primary research literature published since 2008 that primarily focused on early detection of fallopian tube-derived ovarian cancers.

Through the literature review, a multitude of biomarkers and strategies were highlighted that have been investigated including nucleic acid, protein, and metabolic biomarkers. While most reported biomarkers alone failed to improve detection compared to CA125, several studies combined the novel biomarker alone and in combinations with CA125/HE4 (e.g., ANXA2, ADA). In most of these cases, the combined test showed an ability to distinguish between FTE and malignancy that trends towards marginal improvement over single marker testing.

As noted through our literature search, 27.4% (36 of 131) of the studies have demonstrated diagnostic tests with ≥ 80% specificity or AUC ≥ 0.9. As a reference for future work, Table 2 list these studies with the readout and specific findings for each report. These studies should be used to guide future biomarker development.

As the research effort to detect HGSC as a pre-cursor lesion or as an early-stage disease continues, the authors would like to highlight, in the context of this systematic review, areas of potential future research that may aid in the development of a clinically meaningful diagnostic test. These areas include expanded use of proximal fluid, serial sampling, upscaling of a diagnostic test, and improvement in clinical uptake.

Frequently, HGSC progression and dissemination are independent of hematogenous involvement, which highlights the limited utility of identifying circulating tumor cells or macromolecules as an “early” biomarker. Notably, improved technological sensitivity of detection for blood-based biomarkers will likely overcome this limitation, but a more immediate solution may be the use of proximal fluids, such as uterine lavage (UtL). Several studies have employed UtL in high-risk patients (i.e., BRCA-carriers) paired with proteomics or genomics techniques to detect ovarian cancer with mixed results in detecting early-stage disease (35–37). Examination of biomarkers from UtL may represent an approach to overcome a key limitation noted by UKCTOCS that stage-shifting from III/IV to I/II was not sufficient to improve overall survival (5), consistently an active and recruiting clinical trial (NCT04794322) is assessing DNA derived from UtL.

The use of CA125 and HE4 are sufficient to detect large-burden, advanced-stage disease, which is evidenced by the fact that following surgical interventions these biomarkers often fall to within “normal” limits despite the subsequent high disease recurrence rates (38, 39). A strategy to address the limited sensitivity is to use serial sampling of CA125 or HE4 compared to a single threshold to inform clinical decisions (40). By extension, the implementation of a novel biomarker should incorporate serial sampling to improve both HGSC detection and longitudinal understanding of HGSC progression. The serial sampling may allow for better understanding of patient baselines and normal physiologic fluctuations. However, the economic effects of serial sampling must be considered, as biomarker testing can add to patient and health system financial toxicity. Another consideration is the likely anxiety and potential downstream effects of small changes in biomarkers leading to ultimately unnecessary procedures for non-malignant processes.

A noted limitation is that while demographic information, including race, is included, frequently, the data are skewed toward a predominately White cohort that may miss crucial data from minority populations. As it is established that there are differences in CA125 levels among Black and Latinx individuals showing lower baseline CA125 levels than their White counterparts, making it more likely that a diagnosis might be missed or delayed based solely on CA125 levels (41). Additionally, there was an emphasis of early detection of the HGSC subtype using the fallopian tube as the “normal” or healthy comparator, thus there is limited extension of the reported findings to other ovarian cancer histological subtypes. Also, most of the studies are retrospective in nature, which makes it challenging to understand both experimental and selection bias.

Converting novel, research-based diagnostic tests to clinically employed tests represents a significant hurdle due to several factors, including but not limited to manufacturing, development costs, improvement of existing tests, overall low incidence of the disease, health care professional hesitancy, technical training, and data interpretation. As such, these factors need to be addressed with research at a similar level of investment as novel biomarker development itself.

Finally, there would likely be hesitancy from providers to adopt new biomarker testing without evidence of significant improvements in detection rates compared to current standards. Decisions about the clinical implications and a consensus on the interventions to be recommended if there were abnormalities in the new biomarkers will need to be established. For example, there currently is limited evidence and consensus on the management of STIC lesions when detected (42). In parallel with evaluation of new potential biomarkers, it will be critical to generate prospective data that supports management of early-stage disease and detection of precancerous lesions leading to clinically meaningful improvements in outcomes.