Abstract
Ovarian cancer is a leading cause of gynecologic cancer death among women. Tumors diagnosed early (in stage I) have a cure rate approaching 90%. However, because specific symptoms and screening tools are lacking, most ovarian cancers are very advanced when finally diagnosed. CA125 expression and pelvic ultrasonography are of limited efficacy in screening, and the search for new, complementary ovarian cancer biomarkers continues. New technology and research techniques have allowed the identification of over 100 possible tumor markers, many of which are still being evaluated for clinical relevance and several of which have entered clinical trials. Here, we review the methods of biomarker discovery, address the significance and functions of newly identified ovarian cancer tumor markers, and provide further insight into the future of ovarian cancer biomarkers.
Ovarian cancer is the deadliest gynecologic malignancy and a leading cause of cancer death among women. In the USA, an estimated 22,430 new cases will be diagnosed and 15,280 women will die of the disease in 2007 [201]. The cure rate for ovarian cancer when diagnosed in stage I approaches 90%; however, fewer than 25% of tumors are diagnosed in this early stage [1]. Over 50% of ovarian cancer patients survive for 5 years, but the cure rate for those with advanced disease is less than 30%. Therefore, early detection can have a strong effect on survival. Currently, there is no recognized model of screening, since available screening methods lack the sensitivity and specificity required for accurate diagnosis. As a result, many new biomarkers are emerging as possible candidates for use in the diagnosis and treatment response of ovarian cancer.
Biomarkers can be defined as any anatomic, physiologic, biochemical or molecular parameter utilized to measure disease progression or treatment response. They can be detected through physical examination, laboratory assays or medical imaging. The NIH defines a biomarker as ‘a characteristic that is objectively measured and evaluated as an indicator of normal biologic or pathogenic processes or pharmacologic responses to a therapeutic intervention’ [2].
By 2005, an estimated 30 new biomarkers with the potential to be used alone or in conjunction with CA125 had been identified [3]. Since then, over 100 promising candidates have been identified. New studies are being performed to determine potential clinical application and functionality of these markers. In this review, we discuss the challenges and techniques of biomarker discovery and update what has been learned regarding these novel epithelial ovarian cancer biomarkers over the last 2 years.
Screening biomarker criteria
Biomarkers are clinically useful for diagnosing early-stage ovarian cancer, monitoring disease progression and treatment response, and evaluating for disease recurrence. However, to ensure widespread applicability, new diagnostic biomarkers must meet three important criteria [4,5]. First, they should be noninvasive and inexpensive so as to justify their use in screening the general population. This has made especially attractive the search for biomarkers secreted specifically by tumor tissues or immune markers detected in body fluids (e.g., serum or urine). Second, screening biomarkers should be highly sensitive for early-stage disease. Most of the current ovarian cancer biomarkers are not. Third, they must have high specificity and positive predictive value (PPV). Patients who receive false-positive results may be subjected to unnecessary tests and costly and invasive procedures, while those who receive false-negative results may have more advance disease when it is finally diagnosed, thus affecting their prognoses. Therefore, it has been recommended that all new ovarian cancer screening biomarkers should have a minimum PPV of 10% and a minimum specificity of 99.6% [4]. Populations at higher risk, including patients with a strong family history of ovarian or breast cancer, may accept a screening method of lower specificity.
Tools for identifying for ovarian cancer biomarkers
As there is substantial heterogeneity among the various ovarian cancer subtypes, the search for new biomarkers is extremely complicated. However, this complexity has been eased somewhat by the advent of new technology (e.g., in genomics and proteomics) that has allowed the identification of many novel markers over the past few years.
Genomic technology
Genomic technology involves the analysis of whole-genome chromosomal aberrations and the measurement of the expression level of thousands of genes. Multiple technologies have been developed recently for genomic analyses. These include loss of heterozygosity (LOH) analysis, comparative genomic hybridization (CGH), spectrokaryotyping (SKY), serial analysis of gene expression (SAGE), and transcriptome profiling by oligo and cDNA microarrays. Such analyses have facilitated the rapid discovery of the molecular signatures of tumors. Many of these molecular signatures are potential biomarkers. These technologies have been applied to identify genomic changes in ovarian cancer and some of them are summarized as follows.
Loss of heterozygosity analysis
LOH occurs when aneuploidy is demonstrated as an allelic imbalance at the molecular level (e.g., when one parental allele is lost in normal cells) [6]. The allelic imbalance may affect cell-cycle checkpoints and centrosome numbers or may increase the frequency of structural chromosomal changes secondary to chromosome breakage and fusion [7,8]. LOH analysis has provided the opportunity to determine possible locations of critical tumor-suppressor genes and to identify possible cancer biomarkers [8,9]. In LOH studies using a panel of microsatellite markers located on different chromosomes, several researchers have identified multiple regions that are deleted in borderline ovarian tumors and invasive epithelial ovarian cancers [9,10].
SNPs are variations in the DNA sequence in a population at a single nucleotide in the human genome. So far, more than 5 million SNP loci have been identified and validated in a study population. SNPs can occur in the coding and noncoding regions of the genome, which may affect subsequent protein expression. Specific SNPs may also predispose an individual towards tumorigenesis or affect drug metabolism and effect. SNP arrays based on the database of known, validated SNP loci have been developed to interrogate the genotype at 500,000 SNP loci, yielding a resolution more than a 1000-times greater than that achieved using microsatellite markers. This use of SNP arrays has led to the high-resolution mapping and detection of SNP loci exhibiting LOH and copy-number changes in ovarian cancer samples [11].
Comparative genomic hybridization analysis
CGH detects gene copy number in the DNA content of tumor cells. This allows for the analysis of gains and losses in the DNA copy sequence, gains potentially signifying the overexpression of tumor-promoting oncogenes and losses potentially signifying the underexpression of tumor-suppressor genes.
The CGH method involves the competitive hybridization of tumor DNA and normal DNA, both labeled with distinct fluorescent molecules, to normal human metaphase chromosome preparations. The relative intensities of both fluorescent colors are compared using epifluorescence microscopy or quantitative image analysis, and chromosomal regions with a change in gene copy number are then identified. Past research utilizing chromosomal CGH has demonstrated that copy-number abnormalities (CNAs) are significantly more frequent in high-grade ovarian cancers than in low-grade tumors [12]. Examples include the under-representation of 11p and 13q and over-representation of 8q and 7p in high-grade tumors and the under-representation of 12p and over-representation of 18p in well-differentiated and moderately differentiated tumors [12]. Correlations between disease stage and CNA frequency and between chemoresistance and CNA frequency have also been evaluated [13,14].
Transcription profiling
Gene-expression profiling is an array-based technique that allows the simultaneous comparison of the expression patterns of thousands of genes and the tissue in which each gene is expressed [15–17]. There are many transcription-profiling methods, including cDNA and oligonucleotide arrays. In cDNA microarrays, the DNAs of targeted genes are imprinted onto coated glass microscope slides [18]. Total RNA from test and reference samples are fluorescently labeled by reverse transcription and hybridized to the clones on the arrays. Data from each hybridization experiment is then viewed as a normalized ratio between the test and reference samples for each of the genes. Oligonucleotide arrays are similar to cDNA microarrays but are synthesized on the basis of sequence information [19]. This eliminates the need for the clones, PCR products and cDNA required to design cDNA microarrays. Oligonucleotide arrays are also designed to include a large set of probes so as to monitor the expression levels of as many genes as possible.
Several researchers have reported the results of microarray analysis of ovarian cancer specimens [20–24]. Potential serum ovarian cancer biomarkers identified using a cDNA microarray system include prostasin and osteopontin, as well as an autoantibody against an epithelial cell antigen (Ep-CAM) [25–27]. Meinhold-Heerlein et al. developed an algorithm to identify secreted proteins encoded on commercial oligonucleotide arrays and applied it to gene-expression profiles of 67 ovarian cancer specimens and nine normal ovarian tissues samples [28]. They discovered that the combination of CA125 with osteopontin or kallikrein-10/matrix metalloproteinase-7 yielded a sensitivity and specificity of 96–98.7 and 99.7–100%, respectively. Expression profiling of gynecologic malignancies of different histological types and organs of origin has revealed that organ and histology contribute equally to gene expression [29,30]. Other profiling studies have revealed transcription profiles that can determine patient prognosis and survival [31,32].
Proteomic technology
The proteome is the entire protein complement expressed by a genome, cell, tissue or organism. Seeking new biomarkers via analysis of the whole proteome is difficult, since proteomes differ between cells and can be constantly altered by genomic and environmental interactions. Owing to the many post-translational modifications that occur within it, the cancer proteome is estimated to include over 1.5 million proteins [33]. Initially, researchers used 2D polyacrylamide gel electrophoresis (PAGE) to probe proteomes for biomarkers, but the technique was time consuming and provided limited information regarding the proteome [34]. Newer broad-spectrum proteomic approaches – mass spectroscopy (MS), NMR spectroscopy and x-ray diffraction – have emerged and are now being incorporated into the proteomic biomarker-discovery process. MS is used to measure and analyze the mass and charge of proteins [35], whereas NMR spectroscopy and x-ray diffraction are used to evaluate the 3D structure of proteins.
These newer proteomic techniques help identify the components of the cancer proteome and allow comparison of protein expression in normal versus malignant tissues in various stages of disease and in response to therapeutic intervention. The structural knowledge of proteins gained through NMR spectroscopy and x-ray diffraction analysis can illustrate how the proteins interact with other molecules, catalyze enzymatic reactions and are affected by oncogenic mutations. In addition, they allow insights to be made into how proteins are post-translationally modified and the effect that these alterations have on protein function.
Gel-based proteomics
2D-PAGE has been the foundation of proteomic biomarker identification, being used primarily to identify and compare protein expression in normal versus tumor tissues [36]. Over 1000 proteins can be resolved per gel and separated on the basis of charge and size [37]. The desired proteins can then be excised, purified and identified. However, as stated previously, the 2D-PAGE method is extremely time consuming and limited. All potential biomarkers must be validated and tested prior to clinical application, and the method's sensitivity is severely impaired by its inability to detect low-abundance proteins, hydrophobic proteins and proteins of less than 15 kDa or more than 150 kDa [36,38].
Mass spectroscopy
MS allows the analysis of thousands of complex proteins on a variety of platforms. The technique depends on the ionization of large biomolecules and identifies substances by sorting them according to their mass:charge (m/z) ratios [37]. A protein is first digested with an enzyme, usually trypsin, at specific amino acid sequences. The peptides are energized, usually by either electrospray ionization or laser, and then separated in the spectrometer using a variety of methods depending on the type of spectrometer. The separated peptides are deflected onto a detector that measures the signal intensity of each fragment and the m/z ratios of all peptide ions produced [37]. A mass spectrum is created by graphing the signal intensity, or relative fragment abundance, against the m/z ratio. The peptides are then further fragmented into a second mass spectrum, on which each peptide fragment differs by one amino acid, in order to determine the peptide sequence [37]. The sequence is then used to probe protein databases in order to identify the original protein.
The use of laser to ionize peptides can involve either SELDI-TOF or MALDI-TOF, by which a laser is directed at a protein-binding chip containing many patient samples. This interaction causes adherent proteins to be desorbed and launched as protonated and charged ions. The time of flight of the ionized peptides depends on the m/z ratio and is determined prior to hitting the detector. Again, the detector plate records the signal intensity at a certain m/z ratio to generate a mass spectrum. SELDI-TOF is considered to have the higher throughput of the two laser technologies. Vast amounts of data have been produced from a small amount of starting material and analyzed within a very short period of time [36,39].
MS continues to play a critical role in the identification of new ovarian cancer biomarkers. Elevated serum levels of haptoglobin-α chain and inter-α-trypsin inhibitor heavy-chain H4 have been identified in ovarian cancer patients using this technology [40,41].
Important considerations in using proteomic technology
Use of the proteomic techniques to identify new ovarian cancer biomarkers requires careful planning and careful selection of technologies to be used. An important consideration is sample collection. Serum and plasma are now the most widely used bodily fluids, but urine or saliva have also been used. Another important consideration is the fact that validation of newly discovered biomarkers remains a challenge. ‘Between-laboratory’ and ‘within-laboratory’ reproducibility has been demonstrated recently using SELDI-TOF-MS serum profiling [42]. Validation studies need to ensure the specificity and reproducibility of the biomarkers and should address experiment design and appropriate controls [33]. Specifically, areas such as specimen collection, specimen handling and bioinformatic data analysis require further study and examination.
Bioinformatics
Proteomic pattern diagnostics is a bioinformatics-based means of searching for a discriminating series of proteins or protein patterns that, when evaluated together, help to distinguish between unaffected and affected individuals. In a landmark report, Petricoin et al. described the first pattern-recognition algorithms and their application to ovarian cancer detection [43]. In brief, they generated proteomic spectra using SELDI-TOF, and analyzed serum samples from 50 unaffected individuals and 50 ovarian cancer patients to create a training set of mass spectra. They then applied the resulting diagnostic pattern algorithm to 116 masked samples, thereby yielding 100% sensitivity, 95% specificity and 94% PPV in identifying patients with malignant and benign disease. The bioinformatics approach utilized by Petricoin et al. was a genetic algorithm using artificial-intelligence computer-based models to identify the protein spectra by unsupervised testing and retesting.
Other bioinformatics models have also been used to identify novel biomarkers. Zhu et al. created an algorithm in which all biomarkers whose expression levels differed significantly between affected and unaffected individuals were selected from mass spectra using the random-field theory [44]. The best discriminating pattern was then chosen using the best-subset discriminant analysis method. This method yielded 100% sensitivity and 100% specificity. Logical analysis of data has also been applied to create analytical tests that yield 100% sensitivity and 100% specificity [45].
The difficulty in these approaches is due to the complexity of the serum proteome and peptidome, and as a result, still requires vigorous clinical validation. Many identified potential blood peptide biomarker levels are constantly in flux owing to normal physiological and various nondisease-related factors. Therefore, to minimize bias, specific parameters must be implemented during bioinformatics biomarker discovery [46]. The use of inflammatory controls and benign disease as part of discovery sets is crucial to help minimize fluctuations, and these controls must be matched in every possible epidemiological and physiological parameter. The handling and storage of all specimens must occur in a standardized and rapid manner to prevent further bias. Furthermore, the identification and validation of each candidate biomarker in the panel is extremely labor intensive, requiring repetitive iterations and validations to ensure specificity.
Metabolomics
Cellular metabolites are the final products of the interactions between gene expression, protein expression and the cellular environment. The complex biochemical network of these lipids, small peptides, vitamins, protein cofactors and other metabolic substrates and products that interact with each other and with other biological macromolecules is called the ‘metabolome’ [47]. The concentration and ultimate identities of the many components of a metabolome rest on their genomic and proteomic interactions within the tissues. These metabolites are involved in cell transformation from normal to malignant and are present in detectable levels in body fluids [48]. Unfortunately, using this approach in the search for ovarian cancer biomarkers is severely hampered by the lack of a single analytical platform to measure the metabolites and by the large number of unknown metabolites with dynamic ranges in human serum that must be determined in order to complete metabolite profiling [49]. However, steps are being taken to overcome these obstacles. In January 2007, the Human Metabolome Project completed the first draft of the human metabolome, identifying 2500 metabolites, 1200 drugs and 3500 food components in the human body [50].
Metabolomic analysis involves two steps and uses techniques similar to those used for proteomic analysis. In the first step, global analysis via NMR spectroscopy or MS is used to identify and quantify all possible metabolites in a given biological matrix [51]. Interpretation of the complex datasets generated by these studies, which usually requires visualization software and chemometric and bioinformatic methods [52]. In the second step, the substances associated with these fingerprints are identified and then combined to define a biological or clinical end point [52].
Clinical validation
The final and crucial step prior to clinical biomarker application is validation of the potential biomarkers. The measurement platform is tested for sensitivity and precision, and once proven to be reliable, the test is then introduced into independent clinical studies [46]. The sample size of these studies depends on the level of statistical power desired. This relies on both the performance of the peptide analyte panel in the platform-validation discovery phase and the clinical use of the biomarker [46]. For example, general population screening for ovarian cancer would require a larger sample size for clinical validation compared with biomarkers intended for recurrence or treatment response. The true specificity, sensitivity and positive predictive values of candidate biomarkers are elicited through these clinical validation studies since these do not translate from experimental test populations [46].
Uses for biomarkers
Diagnosis of early ovarian cancer
The identification of a novel biomarker of early ovarian cancer would dramatically improve the ultimate prognosis for these patients. Several different techniques, including pelvic ultrasonography, serum CA125 screening and a combination of the two, have been evaluated as possible candidates [53–60]. However, to date, none of these modalities has achieved the sensitivity, specificity and PPV crucially important for any screening test.
Pelvic ultrasonography has excellent sensitivity for the detection of ovarian lesions. However, its poor specificity and poor diagnostic reproducibility prohibit its use as a primary screening tool [53,55,58]. Serum CA125 screening is useful because serum CA125 levels are elevated in 50% of stage I and 90% of stage II ovarian cancers [57]. Nonetheless, this screening tool's false-positive rate is high because there are many benign gynecologic and nongynecologic conditions that can also result in serum CA125 elevation [54]. The new ‘risk of ovarian cancer’ screening method uses an algorithm that incorporates patient age, absolute CA125 levels and rate of change of CA125 levels. This method has demonstrated 99.8% specificity and 19% PPV for detecting primary invasive ovarian cancer [60]. One of the more important aspects of this method is the rate of change in the levels of the biomarkers: elevation and then decline may indicate a transient condition; stable elevation may indicate a chronic, benign condition; elevation and then a further increase may indicate ovarian cancer [60,61].
Currently, two large, ongoing randomized multicenter trials are evaluating a multimodal screening method for ovarian cancer. Preliminary data from the Prostate, Lung, Colorectal and Ovarian Cancer Screening trial shows that an abnormal serum CA125 test has a 3.7% PPV for invasive cancer, an abnormal transvaginal ultrasound has a 1.0% PPV, and the combination of both has a 23.5% PPV [59]. Further studies are being conducted to determine the efficacy of this approach in lowering ovarian cancer mortality. The other screening trial is the UK Collaborative Trial of Ovarian Cancer Screening, which has recruited 200,000 postmenopausal women aged between 50 and 74 years and randomized them to a control group or annual serum CA125 or transvaginal ultrasound screening based on the risk-of-ovarian-cancer algorithm, with the objective of lowering mortality by 30% [202]. As approximately 20% of ovarian cancers express little or no CA125, the use of multiple serum biomarkers in conjunction with these screening methods has been postulated to increase sensitivity and specificity of ovarian cancer screening.
Ovarian cancer screening
Ovarian cancer is an extremely heterogenous disease and thus poses a large challenge to biomarker discovery. There are various histological subtypes of epithelial ovarian cancer (i.e., serous, endometrioid and clear-cell carcinoma), and each is marked by different molecular characteristics and features. As a result, detecting ovarian cancer in its early stages will require a panel of tumor markers. The most comprehensive list of clinically relevant epithelial ovarian cancer biomarkers identified to date is found in Table 1.
Novel potential ovarian cancer biomarkers.
API: α-1-proteinase inhibitor; CASA: Cancer associated serum antigen; CLDN3: Claudin 3; DDR1: Discoidin domain receptor 1; Ep-CAM: Epithelial cell-adhesion molecule; ERCC1: Excision repair cross-complementation group 1; HE4: Human epididymis protein; HIF: Hypoxia-inducible factor; IDO: Indoleamine-2,3-dioxygenase; ITIH4: Inter-α-trypsin inhibitor heavy chain H4; LPA: Lysophosphatidic acid; M-CSF: Macrophage colony-stimulating factor; MDR: Multidrug resistance; MMP: Matrix metalloproteinase; PAI: Plasminogen activator inhibitor; sEGFR: p110 soluble epidermal growth factor receptor; SLPI: Secretory leukoprotease inhibitor; TATI: Tumor-associated trypsin inhibitor; THBS: Thrombospondin.
One study compared the gene expression profiles of 42 ovarian cancers of different histological subtypes with those of five pools of normal ovarian epithelial tissue [62]. Recursive descent partition analysis using one panel of potential tumor markers (i.e., HE4, CA125 and MUC1) could distinguish between tumor and normal ovarian tissues; immunohistochemical analysis using another panel (i.e., CLDN3, CA125, MUC1 and VEGF) could detect all tumors. Another study identified a multivariate model in which a panel of biomarkers, including CA125, apolipoprotein A1, transthyretin and inter-α-trypsin inhibitor heavy-chain H4, had a sensitivity for detecting early-stage ovarian cancer of 74 versus 65% for CA125 alone at a matched specificity of 97% [63]. In yet another study, a combination of CA125-II, CA15-3, CA72-4 and macrophage colony-stimulating factor had a specificity for detecting early-stage disease of 75 versus 48% for CA125-II alone at a fixed specificity of 98% [64].
Mesothelin, a cell-surface protein, has been identified as a potential early diagnostic biomarker in both serum and urine [65,66]. Lysophosphatidic acid, a lipid overproduced in the serum and ascitic fluid of ovarian cancer patients, appears to be a useful diagnostic and prognostic marker [67,68]. Other useful markers, identified in immunohistochemical studies, include kallikrein 10, kallikrein 6, osteopontin and claudin 3 (expressed in 100% of all non-CA125-expressing ovarian cancer specimens), and DF3, VEGF, MUC1, mesothelin, HE4 and CA19-9 (expressed in 29–95% of such specimens) [69]. Of all these potential markers, HE4 is most like CA125 in terms of sensitivity for late-stage disease; in addition, HE4 also appears to have greater specificity for ovarian cancer than for normal tissue and benign disease [69,70].
Treatment response evaluation
Tumor markers are useful for monitoring a patient's response to surgery, chemotherapy or radiation therapy. Often after tumor-reductive surgery, residual disease in the form of small tumor nodules remains and cannot be detected by traditional radiologic methods (e.g., CT, MRI or PET). As a result, serum markers such as CA125 are extremely helpful in determining whether a patient is responding appropriately to treatment. A rising serum CA125 level correlates with disease progression in approximately 90% of cases, and persistently elevated CA125 levels correlate with persistent disease in 95% of cases at the time of second-look laparotomy [71]. Furthermore, patients whose tumors are in remission but ultimately recur often have an elevated CA125 level prior to any other evidence of disease recurrence [71]. However, this is not true for all patients with recurrent disease, and some patients who have normal CA125 levels after treatment may still have persistent small-volume disease. In addition, no studies have demonstrated that early treatment in response to a rising CA125 level in the absence of symptoms improves patient survival [72].
Expression of tumor markers could be useful end points in clinical trial design. For example, investigators could use the levels of serum CA125 to determine whether patients are responding to a particular study treatment. In 19 Phase II trials of 14 different cytotoxic drugs, a 50–75% decrease in CA125 level correlated with response rate [73]. At present, approximately 70% of ovarian cancer patients are sensitive to platinum-based chemotherapeutics, whether administered with or without a taxane drug [71]. New potential biomarkers of platinum resistance include p53, ERCC1, copper transporters and XIAP [74–77]. Several biomarkers have also been identified for taxane resistance, such as MDR1 and tubulin mutations [78,79]. Other potential genetic predictors of sensitivity to platinum-based drugs are MYO5C, SPINK1, ARMCX3, PLEK2 and PRSS11, and for taxanes, TNFSF13B, IFIT3 and BTN3A2 [80]. Further research is needed to determine the clinical applicability of these new gene markers.
Prognosis
Biomarkers can also be utilized to help predict a patient's ultimate prognosis. CA125 half-life appears to correlate with prognosis [81]. For example, the CA125 half-life in patients whose disease is completely resected is 6–14 days; in those with residual disease, the half-life is longer [71,72]. A response to chemotherapy treatment will also shorten the CA125 half-life. Poor prognostic factors include a half-life of greater than 20 days during chemotherapy and failure of CA125 levels to normalize after three cycles of chemotherapy.
The macrophage colony-stimulating factor CSF-1 has been shown to correlate with ovarian cancer cell invasion and to be associated with poor prognosis when present at elevated levels in serum and ascitic fluid [82–84]. Other already identified biomarkers, such as nuclear overexpression of survivin, can indicate improved outcome of ovarian cancer patients [85]. Meanwhile, the prognostic value of other markers (e.g., Her2/neu) remains controversial and will require validation via further research and investigation [86,87].
Conclusion
The complexity of ovarian cancer has made the discovery of novel ovarian cancer biomarkers a challenging and rigorous task. Nonetheless, such tumor markers are important for their potential uses in early disease detection and screening, treatment-response evaluation and planning, and prognosis. New technologies and methods have dramatically changed the way that these biomarkers are sought out. Genomic technologies now allow gene expression, association, regulation and function to be systematically evaluated for possible involvement in tumorigenesis. Proteomic studies of protein expression, structure and function have revealed the complexity of the human proteome due to post-translational modifications and molecular interactions. By combining genomic and proteomic data, metabolomics has opened a new window into the complex biochemical network of macromolecules and metabolites and their possible roles in ovarian cancer.
The prognosis for ovarian cancer is good, provided tumors are found early. Unfortunately, most ovarian tumors are only diagnosed after they have become well advanced, mainly because the symptoms of early disease are vague, and as a result have a much poorer prognosis. Therefore, the search for early screening tools is urgent. The best-studied ovarian cancer biomarkers to date – serum CA125, pelvic ultrasound and combination of the two – still lack the specificity, sensitivity and PPV necessary for clinical implementation. Furthermore, the heterogeneity of ovarian cancer in its various forms, stages and histopathologic subtypes prevents the use of a solitary tumor marker and, in fact, necessitates the use of a panel of biomarkers. Biomarkers such as inhibin A, LDH, AFP and β-HCG are currently used to distinguish and follow nonepithelial ovarian neoplasms, including granulosa cell and other sexcord/stromal cell tumors and ovarian germ cell tumors. However, many of these patients do not express elevated levels of these biomarkers, and despite the rarity of these diseases, more-specific tumor markers are needed to monitor disease incidence and progression [88,89].
Many new biomarkers have been identified for epithelial and nonepithelial ovarian cancers, and several trials are now assessing their clinical relevance when combined with serum CA125 for the early diagnosis of the disease. Fortunately, the list continues to expand as new genes and proteins are discovered and assessed for their potential as diagnostic and prognostic biomarkers.
Future perspective
Advances in genomic technologies have led to the discovery of many promising biomarkers for use in screening and predicting the treatment response and outcomes of ovarian cancer. On the horizon are emerging technologies (e.g., medical resequencing) that will further enhance the ability to predict cancer risk and drug sensitivity. However, because a positive test could lead to surgery, the PPV of any proposed panel of biomarkers would need to be as high as possible to minimize the risk of unnecessary surgery associated with a false-positive result. As greater numbers of tumor samples are screened against such panels, the power to validate their utility will increase, along with the chances that novel molecular targets will be identified and less toxic therapies will be developed. It is hoped that, by revealing the genetic origins of ovarian cancer, the findings of genomic, proteomic, bioinformatic and metabolomic studies will translate into earlier diagnosis and better patient outcomes. Despite some successes, much more research and collaborative effort are needed to identify the proper combination of ovarian cancer biomarkers required for early detection and to resolve conflicting data regarding the utility of several of the newer biomarkers (e.g., secretory leukoprotease inhibitor [28,90] and OVX1 [91,92]). In addition, the sensitivities of all candidate biomarkers need to be further evaluated, since all previous estimates are based on the analysis of preoperative samples obtained from patients who had clinically detectable early- or late-stage disease. The best hope of all is for a simple blood test for early-stage (and therefore highly curable) ovarian cancer that would be available in the near future.
Executive summary
Screening biomarker criteria
Novel diagnostic biomarkers should be noninvasive and inexpensive and have a high sensitivity for early-stage disease, with a high specificity and positive predictive value.
Any new ovarian cancer-screening strategy should have a minimum positive predictive value of 10% and a minimum specificity of 99.6%. Higher risk populations (e.g., individuals with a strong family history of ovarian or breast cancer) may be screened using a method of lower specificity.
Tools for identifying ovarian cancer biomarkers
Biomarkers are being identified using new technologies in the fields of genomics, proteomics, bioinformatics and metabolomics.
Genomics allows the discovery of novel gene biomarkers through the systematic evaluation of gene information and association via loss of heterozygosity analysis, comparative genomic hybridization and transcription profiling.
Proteomics allows the discovery of novel biomarkers through the analysis of protein structure, function and composition via mass spectroscopy and other technologies. Bioinformatics allow affected and unaffected patients to be distinguished through the identification of distinct protein series or patterns.
Metabolomics allows the discovery of potential biomarkers through the evaluation of macromolecules and cell metabolites.
Clinical validation of all new potential biomarkers is crucial to ensure sensitivity and precision of the measurement platform.
Uses for biomarkers: diagnosis of early cancer
Pelvic ultrasonography has a high sensitivity for ovarian masses but has low specificity and diagnostic reproducibility in ovarian cancer detection.
Serum CA125 is elevated in 50% of stage I and 90% of stage II ovarian cancers but can be falsely positive in patients with benign disease.
Large, randomized multicenter trials, including the Prostate, Lung, Colorectal, and Ovarian Cancer Screening trial and the UK Collaborative Trial of Ovarian Cancer Screening, are evaluating potential multimodal screening methods for ovarian cancer using a combination of CA125 levels, transvaginal ultrasound and ‘risk of ovarian cancer’ screening tools.
Uses for biomarkers: ovarian cancer screening
The heterogeneity of ovarian cancer will likely require the use of a panel of biomarkers for screening early-stage disease.
Several studies have been performed that indicate that a combination of CA125 with other biomarkers, such as CA15-3, CA72-4 and apolipoprotein A1, increases the sensitivity and specificity of detecting early-stage disease compared with CA125 alone.
Uses for biomarkers: treatment response evaluation
Elevated CA125 levels are often the first sign of disease recurrence prior to any other evidence of tumor burden.
Biomarkers are useful to monitor disease response to treatment since they often measure tumor burden, which cannot be detected by traditional radiologic modalities.
Several new potential biomarkers have been identified to predict tumor sensitivity to various chemotherapeutic agents, including taxanes and platinum-based drugs.
Uses for biomarkers: prognosis
Biomarkers can be used to correlate and predict patient prognosis. Increased half-life of CA125 levels and macrophage colony-stimulating factor appear to correlate with poor prognosis.
Clinical validation of other biomarkers, including survivin and Her2/neu, are needed to determine their clinical applicability to indicate patient prognosis.
Future perspective
New emerging techniques (e.g., medical resequencing) will further aid in predicting cancer risk and disease sensitivity to various chemotherapeutic agents.
More research and collaborative efforts are needed to identify the proper combination of ovarian cancer biomarkers and to resolve conflicting data regarding some of the newly identified biomarkers.
Footnotes
The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.
No writing assistance was utilized in the production of this manuscript.