Level of evidence: 5
Initially, mass spectroscopy of serum proteins was combined with complex analytic algorithms to identify protein patterns that might distinguish between ovarian cancer cases and controls. This approach assumed that pattern recognition alone would be sufficient to permit such discrimination, and that identification of the specific proteins responsible for the patterns identified was not required. Subsequently, this strategy has been modified, using similar laboratory tools, to identify finite numbers of specific known serum markers that may be used in place of, or in conjunction with, CA 125 measurements for the early detection of cancer. These studies [140,143] have generally been small case-control studies that are limited by sample size and the number of early-stage cancer cases included. Further evaluation is needed to determine whether any additional markers identified in this fashion have clinical utility for the early detection of ovarian cancer in the unselected clinical population of interest.
Level of evidence: 5
Because individual biomarkers have not met the criteria for an effective screening test, it has been suggested that it may be necessary to combine multiple ovarian cancer biomarkers in order to obtain satisfactory screening test results. This strategy was employed to quantitatively analyze six serum biomarkers (leptin, prolactin, osteopontin, insulin-like growth factor II, macrophage inhibitory factor, and CA 125), using a multiplex, bead-based platform. A similar assay was available commercially under the trade name OvaSure? until its voluntary withdrawal from the market by the manufacturer.[Response to FDA Warning Letter]
The cases in this study were newly-diagnosed ovarian cancer patients who had blood collected just prior to surgery: 36 were stage I and II; 120 were stage III and IV. The controls were healthy age-matched individuals who had not developed ovarian cancer within 6 months of blood draw. Neither cases nor controls in this study were well characterized regarding their familial and or genetic risk status, but they have been suggested to comprise a high-risk population. First, 181 controls and 113 ovarian cancer cases were tested to determine the initial panel of biomarkers that best discriminated between cases and controls (training set). The resulting panel was applied to an additional 181 controls and 43 ovarian cancer cases (test set). Pooling both early- and late-stage ovarian cancer across the combined training and test sets, performance characteristics were reported as a sensitivity of 95.3% and a specificity of 99.4%, with a PPV of 99.3% and a negative predictive value of 99.2%, using a formula that assumed an ovarian cancer prevalence of about 50%, as seen in the highly selected research population.
In order to avoid biases which may make test performance appear to be better than it really is, it is worth noting that combining training and test populations in analyses of this sort is generally not recommended. The most appropriate prevalence to use is the disease prevalence in the unselected population to be screened. The prevalence of ovarian cancer in the general population is 1 in 2,500. In a recently published correction to their manuscript, the authors assumed that the prevalence of ovarian cancer in the screened population was 1/2,500 (0.04%) and recalculated the PPV to be only 6.5%. On that basis the investigators have retracted their claim that this test is suitable for population screening. If this test were used in patients at increased risk of ovarian cancer, the actual prevalence in such a target population is likely to be higher than that observed in the general population, but well below the assumed 50% figure used in the published analysis. This revised PPV of 6.5% indicates that approximately 1 in 15 women with a positive test would in fact have ovarian cancer, and only a fraction of those with ovarian cancer would be stages I or II. The remaining 14 positive tests would represent false-positives, and these women would be at risk of exposure to needless anxiety and potentially morbid diagnostic procedures, including bilateral salpingo-oophorectomy.