When ovarian cancer has not spread beyond the ovaries (stage I), up to 90% of patients can be cured with currently available surgery and chemotherapy. By contrast, disease that has spread from the pelvis (stages III-IV) can be cured in only 30% or less. At present, only 25% of ovarian cancers are diagnosed in stage I. Detection of a larger fraction of patients in stage I might impact favorably on survival.
Given the prevalence of ovarian cancer, there are stringent requirements for an effective screening strategy. As diagnosis of ovarian cancer is generally made at surgery, a positive predictive value of 10% implies ten operations for each case of ovarian cancer diagnosed. To achieve a positive predictive value of 10% with a prevalence of 1 in 2,500 requires a high sensitivity of 75% or greater for early-stage disease and a very high specificity of 99.6%.
Approaches to Screening for Epithelial Ovarian Cancer
Three approaches have been utilized for early detection of ovarian cancer: transvaginal sonography (TVS), serum tumor markers, and a two-phase strategy where an abnormal blood test triggers TVS. Over 70,000 women have been evaluated with TVS alone in three large screening trials in the Japan, the United States, and the United Kingdom (Bourne et al. 1993; Van Nagell et al. 2000; Sato et al. 2001). In prevalence screens, sensitivity for stage I ovarian cancer did not exceed 90% (Bast et al. 2002). Specificity was at the margin of that required to achieve a positive predictive value of 10%, with the most promising results observed in the United States trial (Van Nagell et al. 2000). In the United States, however, the cost of annual screening for all women over the age of 50 would be prohibitive. Annual blood tests are potentially less expensive, provided that they were sufficiently sensitive and specific. Among the circulating tumor markers for ovarian cancer, CA125 has been evaluated most extensively. CA125 levels have been elevated 10-60 months prior to conventional diagnosis (Bast et al. 2002). In sera from patients with stage I disease found at conventional diagnosis, CA125 is elevated in 50%-60% (Bast et al. 2002). Specificity of a single CA125 determination in apparently healthy women is 99%, but this falls short of the 99.6% specificity required to achieve a positive predictive value of 10%. Specificity of CA125 can, however, be improved by combining CA125 with ultrasound (Jacobs et al. 1999) or by sequential monitoring over time (Skates et al. 1995).
In the United Kingdom, a randomized trial (Jacobs et al. 1995) compared screening with conventional physical examination (10,985 women) to screening with CA125 followed by transabdominal ultrasound (TAU) if CA125 levels were elevated (10,977). When TAU was abnormal, surgery was undertaken. Twenty-nine operations were performed to detect six cancers, yielding a positive predictive value of 21%, or five operations for each case of ovarian cancer detected. Median survival in the screened group (72.9 months) was significantly greater (p=0.0112) than in the control group (41.8 months).
The trend of CA125 has proven useful in distinguishing malignant from benign disease (Skates et al. 1995). Rising CA125 values are associated with ovarian cancer, presumably related to progressive growth of the source of the antigen. Stable CA125 values, even when elevated, are associated with benign conditions. Steven Skates has developed a computer algorithm that estimates risk of ovarian cancer based on change point analysis. Using changes in CA125 over time to trigger TVS in screening 6,532 women over the age of 50 produced a specificity of 99.8% and a positive predictive value of 19% (Menon et al. 2005). Based on the results of this preliminary study, a large randomized trial (UKCTOCS) was undertaken. A total of 200,000 postmenopausal women have been randomized to three groups: 100,000 controls are followed with conventional pelvic examination annually; 50,000 undergo annual TVS; and 50,000 are monitored with annual CA125 with TVS performed if the risk of ovarian cancer is sufficiently high as judged by the Skates algorithm. Women will be followed for 7 years to determine whether screening improves survival.
Increasing the Sensitivity of Two-Stage Screening Strategies for Ovarian Cancer
Regardless of the outcome of the UKCTOCS trial, CA125 is not likely to provide an optimal initial step in a two-stage screening strategy. At the time of conventional diagnosis, CA125 levels exceed 35 U/ml in 50%-60% of patients with stage I ovarian cancer (Bast et al. 2002). Using change point analysis, the sensitivity of CA125 might be improved by detecting an increase in antigen levels within the normal range. At best, however, the sensitivity of CA125 for early-stage disease is not likely to exceed 80%, as 20% of ovarian cancers express little or no CA125. Greater sensitivity might be achieved with multiple markers, provided that specificity is not compromised. More than 30 markers have been evaluated in combination with CA 125. In these studies, markers have generally been evaluated only two or three at a time, and increased sensitivity has been associated with decreased specificity (Bast et al. 2002).
Over the last 5 years, several different approaches have been utilized to discover potential markers for ovarian cancer. Mesothelin was recognized by the development of murine monoclonal antibodies against tumor associated antigens (Mcintosh et al. 2004). Analysis of growth stimulatory lipids in ovarian cancer ascites detected lysophospha-tidic acid (Xu et al. 1998). Biochemical analysis has documented decreased expression of soluble epidermal growth factor receptor (Baron et al. 2005). Gene expression arrays have detected up-regulation of HE4 (Schummer et al. 1999), kal-likrein 6 and 10 (Diamandis et al. 2000; Luo et al. 2001), prostasin (Mok et al. 2001), osteopontin (Kim et al. 2002), vascular endothelial growth factor (Lu et al. 2004), and interleukin-8 (Lu et al. 2004). Proteomic analysis has detected peptides that are differentially expressed in serum from ovarian cancer patients and healthy individuals. Investigators have attempted to identify a distinctive pattern of peptide expression in serum or, alternatively, to identify specific peptides and to develop individual assays that can be analyzed in combination with other known markers.
Petricoin et al. generated proteomic spectra using SELDI (surface-enhanced laser desorption and ionization) mass spectroscopy (Petricoin et al. 2002). Sera from 50 healthy women and 50 patients with ovarian cancer were compared using an iterative searching algorithm. A pattern that distinguished ovarian cancer sera was then used to classify serum samples from 66 healthy women and 50 women with ovarian cancer, including 18 with stage I disease. All cancers were correctly classified (93%-100%), as were 95% of 66 healthy individuals (87%-99%). While this is an encouraging preliminary study, few patients with early-stage disease were included. Others investigators have reported difficulty in reproducing the analysis from the primary data (Bag-gerly et al. 2005).
A second approach has utilized proteomic techniques to identify specific peptides and to develop individual assays that can be analyzed in combination with conventional markers. Zhang et al. utilized sera from five different academic centers to identify upregulation or downregula-tion of peptides that could be found in all five data sets (Zhang et al. 2004). Three biomarkers were identified: apolipoprotein A1, a truncated form of transthyretin, and a fragment of inter-alpha-1-trypsin inhibitor heavy chain 4 (IATI-H4). Changes in expression of the three markers added 9% sensitivity to CA 125II at a constant specificity of 97%. Interestingly, the IATI-H4 fragment is flanked by potential tissue kallikrein cleavage sites and increased expression of kalli-kreins has been documented in ovarian cancers (Diamandis et al. 2000; Luo et al. 2001). Thus, the fragment might be generated by proteolytic cleavage by tumor derived kallikreins as normal plasma components percolate through the ovarian cancer stroma, which would provide a strong biologic basis for this marker.
Identification of an optimal combination of biomarkers for early detection of ovarian cancer will require analysis of multiple assays on a common panel of sera from healthy individuals and from ovarian cancer patients at the time of conventional diagnosis. Data of even greater importance must be obtained from prediagnos-tic samples collected during screening trials from patients destined to develop ovarian cancer. Such serum specimens have generally been preserved in relatively small amounts. Consequently, the simultaneous assay of multiple markers in small volumes of sera will be important for the identification of optimal panels. Luminex technology permits simultaneous multiplexing of several assays in small volumes of serum (Vignali 2000). Sets of beads are marked with different concentrations of a red fluor. Double determinant assays are constructed with primary antibodies reactive with different markers on different sets of beads. Second antibodies reactive with each marker are labeled with a single green fluor. Flow cytometric analysis can permit simultaneous analysis of multiple markers in as little as 50 |L of serum. Promising preliminary data have been obtained by Anna Lokshin at the University of Pittsburgh (Gorelik et al. 2005). Resent unpublished observations from this group indicate that the multimarker approach can achieve a sensitivity of 90%-92% at a specificity of 98%.
To increase sensitivity without losing specificity using multiple markers, sophisticated statistical techniques are required. In two recent studies (Skates et al. 2002, Zhang et al., in revision), CA125II, CA72-4, CA15-3, and M-CSF were measured in a validation set of 60 early-stage ovarian cancer and 98 control sera. At a specificity of 98% that would require TVS in only 2% of patients in a two-stage screening strategy, CA125II exhibited a sensitivity of 48%. At the same level of specificity, artificial neural network analysis with the four markers produced a sensitivity of 72% (Zhang et al., in revision) and a mixture of multivariate normal distributions produced a sensitivity of 75% (Skates et al. 2004). Consequently sensitivity could be increased by 24%-27% without a loss of specificity at the level required for a two-stage screening strategy.
Was this article helpful?
Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.