Abstract
Over 90% of women diagnosed with localized ovarian cancer survive at least 5 years, but because the disease is often advanced at diagnosis, only half of patients survive that long. Currently, screening is only recommended for women with a strong family history of the disease, for whom prophylactic surgery is appropriate once childbearing is complete. Screening is not recommended for the general population since currently available screening tests seldom detect curable tumors, and often lead to unnecessary surgery. New markers are under investigation that may signal disease earlier and be useful in a marker panel to select women for imaging and/or surgery. A clinical challenge is that definitive diagnosis requires major abdominal surgery, and screening tests often detect benign ovarian conditions. A challenge for translational research in biomarkers for early detection is that access is needed to preclinical blood samples, obtained some months prior to diagnosis, in order to determine if candidate markers can detect disease prior to onset of symptoms.
Progress in translational ovarian cancer research can be measured by improvement in outcomes for women with, or at risk of, ovarian cancer. From a clinical perspective, for women diagnosed with ovarian cancer, relevant outcomes are survival and quality of life. From a public health perspective, for all women at risk of ovarian cancer, relevant outcomes are incidence and mortality rates.
Screening is designed to improve survival and quality of life of cancer patients through early diagnosis and treatment; its efficacy is generally evaluated in terms of reduction in mortality. As ovarian cancer often remains asymptomatic until it is well advanced, there is an opportunity for significant improvement through early detection. During the period 1996–2002 in the USA the 5 year survival rate of women diagnosed with ovarian cancer was 45% overall; in women with cancer confined to the ovaries it was 93%, but only 19% of ovarian cancer was diagnosed in this very early stage [101].
While promising and potentially cost-effective [1], detecting ovarian cancer early through screening presents several challenges. The first is that ovarian cancer is a relatively uncommon disease. The age-adjusted annual incidence rate of ovarian cancer in the USA is approximately 14/100,000 women. Among women over the age of 50 years, in order to achieve a positive predictive value (PPV) of 10%, a screening test with 80% sensitivity needs to have 99.6% specificity. Screening tests should not detect common benign ovarian disease, as definitive diagnosis requires surgery that carries significant risks, morbidity and costs; it may also lead to unnecessary removal of gynecological organs. Second, since only 7–10% of women with ovarian cancer have a high-risk family history, limiting screening to these high-risk women cannot significantly impact overall ovarian cancer mortality. Third, the natural history of ovarian cancer development and progression is largely unknown, and its precursor lesion has not been identified. Finally, although both a serum marker and an imaging modality are currently available, neither has been demonstrated to detect ovarian cancer early enough to enable cure, and neither has adequate specificity to avoid unnecessary surgery.
The best screening tests detect cancer before it becomes invasive, by identifying precursor lesions. Since 1950, the Pap smear has reduced the incidence of, as well as the mortality from, invasive cervical cancer in the USA by 70% [2]. Similarly, screening by colonoscopy can prevent invasive colon cancer through identification and treatment of premalignant polyps [3], and mammography screening identifies ductal carcinoma in situ (DCIS) that might progress to invasive breast cancer in the absence of treatment [4]. When precursors are unknown or undetectable, screening for elevated risk can be useful to reduce disease incidence. For example, screening for and treating high cholesterol and high blood pressure effectively reduces the incidence of myocardial infarction in individuals at risk. Opportunities for reducing the incidence of ovarian cancer through screening for risk markers may be important to explore due to the many challenges to early detection of curable invasive lesions. Research to identify ovarian cancer precursors and risk markers is likely to contribute significantly to the development of a successful ovarian cancer screening strategy.
Current & future developments in screening for ovarian cancer
Ovarian cancer screening: current clinical standards
The National Comprehensive Cancer Network Clinical Practice guidelines recommend screening starting at the age of 35 years, or 5–10 years earlier than the earliest age of ovarian cancer diagnosis in the family, for women at high risk who have not elected to undergo risk-reducing salpingo-oophorectomy (RRSO; bilateral prophylactic removal of the ovaries and fallopian tubes) [5,102]. A standard screening protocol has not been defined but transvaginal sonography (TVS) and the serum marker cancer antigen 125 (CA125) are often used every 6–12 months. Screening is not recommended for average-risk women.
A woman is considered high risk if, after careful risk assessment including a detailed personal and family cancer history and/or genetic testing results, she meets inclusion criteria for the Hereditary Breast and Ovarian Cancer Syndrome [102]. Inherited mutations in BRCA1 are associated with an estimated lifetime risk of breast cancer as high as 90%, as well as a 40–50% risk of ovarian cancer; ovarian cancer risk associated with BRCA2 mutations is lower but substantial at 20–30% [6]. Hereditary Nonpolyposis Colorectal Cancer (HNPCC) is a genetic syndrome caused by germ-line mutations in a family of genes involved in DNA mismatch repair. The same mutations also increase ovarian cancer risk; affected women have a lifetime ovarian cancer risk approaching 12% [6].
It is recommended that all women with known BRCA1 or BRCA2 mutations or women with a strong family history of breast and ovarian cancer be offered RRSO once child-bearing is complete [7]. Surgery can usually be performed laparoscopically with little surgical morbidity [8]. Careful analysis of the RRSO specimen is mandatory to identify the 3–12% of patients with occult invasive cancer at the time of surgery [5,9,10]. When performed prior to menopause, RRSO also reduces the risk of developing breast cancer by approximately 50% [11,12]. Optimal timing of RRSO is unclear, as the risk of ovarian cancer begins to rise at the age of approximately 40 years in BRCA1 mutation carriers but not until the age of 60 years for BRCA2 carriers [13]. Delaying surgery until near the time of natural menopause may limit the protection against breast cancer [11].
Ovarian cancer screening: evidence from research
Screening tests evaluated in prospective studies
Two types of ovarian cancer early detection tests are currently under investigation: imaging and blood tests. Imaging using TVS appeared promising in early studies, but it identifies benign as well as malignant masses. Estimates of the specificity of TVS when used as a first-line screen in the postmenopausal population have been reported from the Prostate, Lung, Colon and Ovary (PLCO) trial. Of nearly 29,000 women screened at the first visit, 4.7% had an abnormal TVS. The PPV of TVS for invasive cancer was 1% [14], meaning that only 1 of 100 positive tests yielded a cancer at surgery. DePriest and his colleagues at the University of Kentucky have been working to improve the specificity of TVS using a morphology index [15]. This group screened 14,469 predominantly average-risk postmenopausal women with annual TVS [16]. Of 180 women who underwent surgery, 17 were found to have ovarian malignancy, for a PPV of 9.4%. Four women developed ovarian cancer within 12 months of a negative screen, for a sensitivity of 81%.
Despite frequent screening in the high-risk population, when TVS detects ovarian malignancy the disease is often advanced. Fishman screened 4526 women with a family or personal history of breast or ovarian cancer using TVS every 6 months. Among 98 women with persistent pelvic masses, 49 required surgery. Two epithelial ovarian cancers, four primary peritoneal carcinoma, four fallopian tube carcinoma, two endometrial adenocarcinoma and 37 benign ovarian tumors were identified. The two endometrial carcinoma were stage IA, but the ovarian, fallopian tube and primary peritoneal carcinoma were all advanced (stage III) [17].
Blood tests are potentially very useful in an ovarian cancer screening strategy. CA125, a mucin-like glycoprotein, is elevated in the serum of most women with ovarian cancer but its performance has not justified its use as a stand-alone screen. Using a cut-off of 35 U/ml, the specificity of CA125 is below 95%, and sensitivity in early stage disease is poor. Preoperative serum levels of CA125 are below 35 U/ml in roughly 50% of clinically detected stage I cases [18] and in the majority of women with occult cancers identified at RRSO [19].
Multimodal screening strategies evaluated prospectively
The use of CA125 and TVS together in a multimodal screening strategy has been explored extensively. Reports suggest that sensitivity for early stage disease is limited in the high-risk population. Hogg reviewed findings from 12 studies using CA125 and TVS to screen over 6000 high-risk women [20]. Excluding borderline and germ cell tumors, there were 38 ovarian cancers identified, only nine of which were stage I; 15 cancers diagnosed within a year of a screen were missed by both CA125 and TVS. Similarly, Sterling identified two stage I invasive cancers among 12 ovarian cancers detected in a cohort of 1100 high-risk women participating in a screening program [21].
To improve sensitivity while maintaining good specificity, the Risk of Ovarian Cancer Algorithm (ROCA) was developed for use in a multimodal screening strategy. In an analysis involving 33,621 CA125 results from 9233 postmenopausal women over the age of 45 years, it improved sensitivity from 62 to 86% at a specificity of 98% [22]. The ROCA uses a change point model to interpret longitudinal CA125 values in the context of other variables including age and menopausal status, in order to stratify women based on their risk for ovarian cancer at the time of the screen. Women are asked to return for repeat CA125 testing and/or ultrasound screening if their CA125 levels are abnormal. In a prospective screening trial of 6682 average-risk postmenopausal women the specificity and PPV for ROCA at the prevalence screen were 99.8 and 19%, respectively [23], a significant improvement in screening performance compared with a single threshold rule, such as above 30 U/ml or 35 U/ml.
Two prospective screening trials targeting high-risk women are currently underway. Neither includes a control arm as it is considered unethical to randomize high-risk women to a nonscreening arm. Both use the ROCA to estimate an individual's risk of ovarian cancer based on serial CA125. The UK Familial Cancer Screening Study is screening over 1500 high-risk women using annual CA125 and TVS testing. In the USA, the Cancer Genetics Network (CGN) ROCA study is screening over 2200 high-risk women at multiple centers. CA125 is measured quarterly to stratify women using ROCA. Women at usual risk return to routine screening, intermediate-risk women are referred for TVS, and elevated-risk women are referred for TVS and subspecialty consultation. TVS is performed annually at some centers.
A multimodal strategy has also been tested in the average-risk population. Jacobs enrolled 21,935 postmenopausal women who had participated in a single prevalence screen in a pilot randomized, controlled trial (RCT) [24]. The RCT had two-arms, a no-screening control group or multimodal screening using a CA125 serum concentration level of over 30 U/ml to select women for imaging by TVS. Women with abnormal CA125 and TVS were referred for surgery. The PPV for this strategy was 20%.
The pilot trial was too small to detect a mortality reduction, but a difference in survival provided the rationale for a large efficacy trial that is currently being conducted in the UK. The design for the efficacy trial includes two intervention arms, as shown in Table 1, to test the multimodal strategy against the more costly approach of screening all women annually with TVS, as well as a no-screening control arm. As in Jacobs' pilot RCT, CA125 is used as a first-line screen in the multimodal arm. To improve sensitivity of the first-line screen with no loss of specificity, the ROCA, described above, is being used to select women for TVS. In the USA, a different multimodal strategy is being tested in the PLCO trial. It is designed to maximize sensitivity in the sense that the screen is considered positive if either CA125 or TVS is abnormal [25]. CA125 is considered normal if it is below 35 U/ml. Design parameters for the ovarian cancer screening component of the PLCO trial are shown in Table 2. The results of both trials will be available within the next decade.
Design parameters for the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS): a randomized, controlled trial.
ROCA: Risk of Ovarian Cancer Algorithm; TVS: Transvaginal sonography.
Design parameters for the ovarian component of the Prostate, Lung, Colon and Ovary (PLCO) randomized, controlled screening trial.
The PLCO design was revised to continue screening for an additional 2 years using only CA125.
CA125: Cancer antigen 125; TVS: Transvaginal sonography.
Meeting the challenges of screening for ovarian cancer
Improving the ability to identify invasive disease: using a marker panel
Technologies emerging in the last decade, including genomics and proteomics, are enabling scientists to identify new ovarian cancer diagnostic markers that may improve sensitivity for detecting early stage disease. The many markers that have been identified have been reviewed elsewhere [26]. To date, efforts to discover, develop and validate candidate markers have been of three general types. The first approach has relied on new proteomics technologies to identify patterns of circulating, unidentified proteins. The second approach has used novel statistical algorithms to combine or use previously identified markers in new ways. A third approach has been to identify novel markers and evaluate them for their contribution to a panel that includes CA125.
The first approach made its debut 4 years ago, when it was reported that a proteomic pattern interpreted by a bioinformatics program could distinguish women with ovarian cancer from healthy women with 100% sensitivity, 95% specificity, and 94% PPV [27]. The PPV must be interpreted with caution, as it reflects the prevalence of cases in the study rather than in the population. Discovery of the pattern was performed using serum samples from women with and without ovarian cancer, generating excitement in the research as well as the advocacy communities. However, the pattern involved analysis of many unidentified peptides and proteins and has proved difficult to replicate or validate. Methods that allow identification of the proteins that contribute to the pattern may yield more reproducible results [28]. More recently, a glycosylated form of eosinophil derived neurotoxin (EDN) and a cluster of COOH-terminal osteopontin were identified in urine from women with ovarian cancer using proteomic-based discovery techniques [29]. Enzyme-linked immunosorbent assays (ELISAs) were developed and the markers were measured in pretreatment urine samples from 128 women with ovarian cancer, 52 women with benign conditions, 44 women with other cancers and 188 healthy controls. Performance of the marker combination in early stage ovarian cancer was 72% sensitivity at 93% specificity.
The second approach is perhaps best demonstrated by work published last year in which four known serum markers (prolactin, insulin growth factor II [IGF-II], osteopontin and leptin) were selected from among 169 candidate markers to be included in a panel to distinguish women with ovarian cancer from healthy women [30]. Results reported were comparable to those achieved by the proteomic patterns described above, with sensitivity, specificity and PPV all achieving 95%; similar caution is needed in interpreting the PPV. Recently, a second example of the approach uses circulating autoantibodies, rather than proteins [31]. Evaluation of autoantibodies in serum from women with early stage as well as advanced ovarian cancer and healthy women yielded sensitivity and specificity of 86% and 92%, respectively; at 100% specificity, sensitivity was 56%. Like proteomic pattern detection, this approach relies on training and test sets to develop and evaluate a panel. Independent replication of these studies followed by validation in preclinical samples is needed to confirm the value of these marker panels for early detection.
The third approach seeks to improve on existing screening methods by adding known or novel markers to a panel that includes CA125. A panel consisting of CA125 and known tumor markers CA 72–4 and macrophage colony-stimulating factor (M-CSF) increased the sensitivity, at 98% specificity, for stage I disease to 70% compared with 40% for CA125 alone [32]. A number of novel candidate markers have emerged including human epididymis protein (HE)4 [33], mesothelin [34], the kallikreins 6, 10 and 11 [35], and B7-H4 [36]. Several of these markers can be detected by immunohistochemical (IHC) analysis in ovarian cancer tissues with little or no expression of CA125 [37]. HE4 has been shown to be more specific than CA125 in discriminating women with malignant tumors from those with benign tumors [33]. Mesothelin in combination with CA125 has been shown to perform better than either used alone in discriminating women with malignant tumors from healthy women [38], and a panel combining HE4 and CA125 (both on a bead-based platform) performs better than either marker used alone [39]. Based on protein expression in tissue, HE4 and Mesothelin have been shown to complement CA125 with good specificity [40].
Increasing lead time: using a longitudinal algorithm
The goal of screening is not to identify symptomatic cancer but to detect asymptomatic cancer early enough that it can be cured. Lead time is defined as the interval between detection by screening and clinical diagnosis based on symptoms. Lead time is a function of the characteristics of the screening tests used, screening frequency, and the decision rule(s) used to select women for definitive diagnostic procedures. A serum marker is most useful as a screening test if it provides signal early in the preclinical phase of the disease. Estimation of preclinical marker behavior, including potential lead time, requires measurement of markers in serial blood samples collected several months to several years prior to clinical diagnosis. A longitudinal algorithm can potentially improve lead time for markers that are relatively stable over time within women (lower variability within than between women), because smaller changes in these markers' levels are needed to distinguish signal from noise. The ROCA uses information regarding the variability in CA125 levels over time in both cases and controls to estimate the risk of cancer. A simpler parametric empirical Bayes (PEB) approach has been proposed for using serial serum marker data that tailors the screening decision rule to the individual woman [41]. The PEB uses information regarding the behavior of a marker over time in healthy, asymptomatic subjects as well as an individual woman's personal screening history to determine her own individual marker cut-off level at each screen. Cut-off levels assigned by the PEB will be lower for most women than a single threshold rule with comparable specificity [42]. Lower cut-off levels translate into longer lead times for screen-detected cancers. Importantly, the PEB can be easily generalized to a marker panel, and can be used for novel markers that have not yet been evaluated in preclinical samples.
Clarifying the goal: identification of invasive versus preinvasive disease
Preclinical samples from large prospective studies such as the Women's Health Initiative (WHI) and PLCO trials are the key to understanding the potential utility of a marker for early detection. These samples are very valuable because they allow estimation of marker behavior prior to the development of the symptoms that lead to diagnosis. However, they cannot reveal the presence or absence of invasive disease at the times prior to diagnosis when the preclinical serum samples were obtained. Accordingly, they are most useful for estimating the relative risk of subsequent ovarian cancer diagnosis based on marker levels or changes in marker levels. Knowledge of the presence or absence of disease at the time a marker first provides signal requires a prospective screening study in which surgical intervention is triggered by the marker.
The only ovarian cancer serum marker for which data from preclinical samples are available is CA125. Using the Janus databank and a nested case–control design, preclinical CA125 levels were estimated for 668 ovarian cancer patients (478 invasive and 190 borderline) and 1989 matched controls. The authors conclude that CA125 was not sensitive enough to be an early detection marker [43]. Nevertheless, CA125 levels were significantly higher in cases than in controls up to 10 years prior to diagnosis, and increased with proximity to diagnosis. It is therefore possible that CA125 might serve as a risk marker for ovarian cancer.
Because they have not yet been evaluated in preclinical samples, the novel markers described above have been shown only to be diagnostic markers. The distinction among diagnostic, early detection, and risk markers is illustrated in Figure 1, which depicts the behavior of three hypothetical markers as cancer progresses through a precursor lesion stage, early and advanced cancer stages, and diagnosis based on symptoms. Markers A, B and C are all elevated at the time of diagnosis, when clinical samples are obtained and the cancer is typically advanced and symptomatic, but they are not equally good early detection markers because they behave differently prior to diagnosis. Marker A performs well as a diagnostic marker because it is highly elevated in women with cancer who present clinically with symptoms, but it is less useful as an early detection marker because it does not elevate until the tumor is quite advanced. Marker B would perform well as an early detection marker because it elevates while the disease is still likely to be curable, but it may signal disease in women who do not yet have invasive cancer, much as mammography detects DCIS as well as invasive disease. Marker C is most useful as a risk marker because it predicts disease in the future, much like a strong family history or a mutation in the BRCA1 or BRCA2 gene. Because the goal of screening is to detect disease that is curable, not just to identify disease, markers of the first type may be inadequate even when used together in a panel.
The signal provided by a screening test prior to symptoms and clinical diagnosis determines its utility as a diagnostic (A), early detection (B), or risk (C) marker.
Preventing invasive cancer: screening strategies for the future
Even if marker panels and longitudinal algorithms improve rates of disease detection, achieving detection early enough to enable cure will remain challenging. A substantial proportion of women diagnosed with advanced-stage, serous ovarian cancer have normal appearing ovaries by TVS as little as 3–12 months prior to clinical diagnosis [44]. Furthermore, careful surgical staging of high-risk women with occult cancer identified at the time of RRSO demonstrates that extraovarian spread can occur even when the primary tumor volume is quite small [9,20]. Therefore it may be necessary to invoke a different screening paradigm, focusing on markers that predict, rather than detect, disease.
A prevention approach – RRSO – has already been identified that reduces ovarian cancer incidence dramatically in women at high risk for the disease. A prospective cohort study of 170 women and a retrospective case–control analysis of 551 women, all with BRCA1 or BRCA2 mutations, both confirm that RRSO substantially reduces ovarian cancer incidence [12,45]. Compared with control women who did not undergo surgery, the relative risk for developing ovarian or peritoneal cancer among women undergoing RRSO was 0.04 (95% confidence interval [CI]: 0.01–0.16) over a mean follow-up of 8.8 years [12]. Prophylactic surgery including bilateral removal of the ovaries and fallopian tubes and hysterectomy also reduces gynecological cancer risk in women with HNPCC mutations [46].
Protection from ovarian cancer is not complete, as a small proportion of women will develop an ovarian-like cancer of the primary peritoneal lining following RRSO [47]. RRSO is also not without risk. Oophorectomy in premenopausal women increases the risk for cardiovascular morbidity, osteoporosis, endocrine-associated symptoms including hot flashes and difficulties with sexual function [48,49]. The relative risks and benefits of hormone therapy or alternative therapies in high-risk women to ameliorate the risks of oophorectomy are incompletely understood, although short-term hormone replacement therapy does not appear detrimental [50].
Currently, RRSO is reserved for women with an inherited susceptibility associated with a 15–50% lifetime risk of ovarian cancer, based on a strong family history or mutations in BRCA1 and BRCA2. If risk markers of acquired susceptibility, analogous to elevated cholesterol or blood pressure for heart disease, are identified that can predict ovarian cancer risk as accurately as BRCA1, they could also be used to select women for RRSO. Markers that could predict 5- or 10-year risk would be particularly useful. The cost and morbidity, and thus the threshold risk level for RRSO, could be reduced if RRSO procedures were performed at the time of abdominal or pelvic surgery for other indications. At-risk women cannot be expected to have invasive disease at the time of RRSO, of course, but it is possible that some would have molecular features associated with premalignant conditions including overexpression of p53 [51] or the novel marker HE4 that is not normally expressed by the epithelial cells found in the surface of the ovary [52].
Conclusion
Current screening strategies may be inadequate in both the high-risk and postmenopausal populations. Frequent screening using a panel of serum markers interpreted by a longitudinal algorithm will probably be needed to detect invasive ovarian cancer early enough to improve outcomes. Surgical removal of the ovaries and fallopian tubes is an effective way to prevent ovarian cancer in women at high risk based on family history. Reduction of ovarian cancer mortality through screening may require new approaches. The use of preclinical samples is the key to both validation and identification of serum markers for use in screening for ovarian cancer. Detection of preinvasive lesions may be an appropriate goal for ovarian cancer screening. In the clinic, marker panels may be useful to identify women for whom risk is sufficiently high that prophylactic surgery is justified.
Future perspective
A new paradigm may be needed to meet the challenges presented by ovarian cancer screening. Despite advances in proteomic and related technologies, it may remain very difficult to identify invasive disease that is curable. A model similar to that used in heart disease may be relevant, in which individuals at high risk are identified and treated prior to disease onset. The emergence of novel technologies that allow the identification of proteins in complex mixtures such as serum, combined with the availability of serum samples from large trials will likely yield a marker panel that can identify women who are destined to be diagnosed with ovarian cancer but do not yet have the disease. Repositories associated with trials such as the PLCO and WHI contain blood samples that could be used to discover such risk markers; PLCO samples could be used to validate markers discovered using WHI samples, and vice versa. Interventions, including salpingo-oophorectomy, in women at increased risk can then potentially reduce the incidence of, as well as mortality from, ovarian cancer. This risk-assessment approach to screening may hold the key to the eradication of ovarian cancer in the future.
Executive summary
Less than 20% of ovarian cancer is diagnosed early, when it is potentially curable. Although both a serum marker and an imaging modality are currently available, neither has been demonstrated to detect ovarian cancer early enough to enable cure, and neither has adequate specificity to avoid unnecessary surgery.
The best screening tests detect cancer before it becomes invasive, by identifying precursor lesions. Research to identify ovarian cancer precursors and risk markers is likely to contribute importantly to the development of a successful ovarian cancer screening strategy.
Screening is not recommended for average-risk women; for women with known BRCA1 or BRCA2 mutations or a high-risk family history, screening is recommended until risk-reducing salpingo-oophorectomy (RRSO) can be performed once child-bearing is complete.
Two types of ovarian cancer early detection tests are currently under investigation: imaging and blood tests. The use of the blood test cancer antigen 125 (CA125) and transvaginal sonography (TVS) together in a multimodal screening strategy is being tested in randomized, controlled trials in the USA and the UK.
A multimodal strategy tested in the average-risk postmenopausal population, using a serum concentration of CA125 over 30 U/ml to select women for imaging by TVS, yielded a PPV of 20% and a significant difference in survival.
Efforts to discover, develop and validate candidate markers involve new proteomics technologies to identify patterns of circulating proteins, novel statistical algorithms to use previously identified markers in new ways, and identification of novel markers that complement CA125. A longitudinal algorithm can potentially improve lead time for markers that are relatively stable over time within women.
Preclinical blood samples are very valuable because they allow estimation of marker behavior prior to the development of the symptoms that lead to diagnosis. It may be necessary to invoke a different screening paradigm focusing on markers that predict, rather than detect, disease in order to reduce ovarian mortality.
Frequent screening using a panel of serum markers interpreted by a longitudinal algorithm may detect some invasive ovarian cancer early enough to improve outcomes. Since surgical removal of the ovaries and fallopian tubes is an effective way to prevent ovarian cancer, detection of preinvasive lesions may be an appropriate goal for ovarian cancer screening.
It may remain very difficult to identify invasive disease that is curable. A risk-assessment approach to screening may hold the key to eradication of ovarian cancer in the future.