Abstract
Menon and colleagues report interim results of a large clinical trial into the early detection of ovarian cancer [1]. There is considerable interest regarding the need for screening programs for the early detection of ovarian cancer. Since the symptoms are vague, the detection of cancer is mostly late and difficult to treat. This is reflected in a 65% detection rate when the cancer has already spread beyond the ovary, with a 5-year survival rate of 30%. Detection of cancer while still confined to the ovary, although less common (25%) increases the survival rate to approximately 90% [2]. Thus, developing procedures to identify ovarian cancers at its earliest stages would save lives.
One common approach to detect cancer at its early stages has been to identify cancer-specific markers in blood or urine. Recent reviews have listed more than 20 proteins that show some evidence of suitable specificity and sensitivity [2,3]. However, since the disease is relatively uncommon, the specificity and sensitivity characteristics of any proposed test must be particularly stringent. The prevalence of ovarian cancer in the general postmenopausal female population is approximately 1:2500 [2]. For a test to be suitably specific, the proportion of cancer cases detected of those positively screened (positive predictive value) should be greater than 10%, which translates into a specificity value for the test of approximately 99.6%, that is to say, a false-positive level of 1:250. Currently, the specificity of the bulk of tests examining single or multiple analytes is 95–98%. In these cases, the proportion of cancer cases detected is 1–2.5% of those screened. This is promising, but not good enough.
The most common blood marker available is CA125, also known as MUC16. In addition to ovarian cancer, the blood levels of this marker are elevated in other clinical conditions, including endometriosis and other peritoneal cancers. However, CA125 demonstrates a high sensitivity (>95%) for the aggressive forms of ovarian epithelial cancer (mostly serous), which are also the most common. Many of the more recently identified markers have been assessed in conjunction with CA125. As a static marker of ovarian cancer it is poor, with a sensitivity of 50–90% depending on the stage of the disease [2]. Skates et al. noted that a higher sensitivity was obtained for CA125 (showing an increase from 70 to 84%) if the rate of change in CA125 serum levels in conjunction with age is used as the predictor rather than a fixed cut-off CA125 value [4]. Subjects with ovarian cancer showed progressive increases with time while healthy subjects remained unchanged. An algorithm describing the similarity in serum profiles observed with controls and cancers was developed so that serum CA125 levels determined over a minimum of 1 year can be classified as normal, intermediate or at high risk of ovarian cancer [4].
Results
In 2001, Jacobs and colleagues undertook a large population-based, randomized trial to establish whether serum CA125 and/or ultrasound observation were suitable as screening tests for the detection of ovarian cancer. In April 2009, an updated report from the first 5 years of the trial (2001–2005) was published in Lancet Oncology [1].
The trial consisted of 202,638 postmenopausal women who were screened annually over 5 years. These women were randomly divided into three groups: a no-screen group (101,359 women) and two screen groups. In one screen group, serum CA125 was determined, followed by an transvaginal ultrasound examination if the serum CA125 pattern was classified as at-risk (multimodal screening [MMS] 50,640 women); the second screening group used ovarian ultrasound (USS; 50,639 women) alone. The ultrasound analyses were described in terms of defined morphological criteria in order to standardize scans across centers [5] with the use of confirmatory scans by those experienced in gynecological scanning. Subjects at high risk were then clinically assessed and, if appropriate, followed-up by surgery.
The sensitivity, specificity and positive predictive values for all primary ovarian and tubal cancers for the MMS group were 89.4, 99.8 and 43.3%, respectively, and for the USS group were 84.9, 98.2 and 5.3%, respectively, with a significant difference in specificity between groups. In the MMS group, 409 women were identified as high risk based on CA125 levels, which decreased to 167 following transvaginal ultrasound and 97 following clinical assessment. Of these, 42 primary ovarian and tubal cancers were detected at surgery. In the USS group, 2785 women were identified with abnormal ultrasound results, 1894 of these women required clinical assessment, 884 women underwent surgery and 45 primary ovarian and tubal cancers were identified. The higher rate of surgery in the USS group is attributed to the presence of benign lesions in the ovary, which are poorly distinguished by ultrasound.
Of the 87 ovarian or tubal malignancies detected in the two screening groups, 28 were borderline epithelial cancers (largely nonaggressive) and 58 were primary invasive epithelial cancers of which 28 were at the earliest stages (stages I/II). More borderline tumors were identified in the USS group (20 patients) compared with the MMS group (two patients). When only primary invasive epithelial ovarian and tubal cancers (excluding borderline tumors) were considered, the sensitivity, specificity and positive predictive values for MMS were similar (89.5, 99.8 and 35.1%, respectively) to those obtained with all primary ovarian and tubal cancers. An additional 13 ovarian cancers were diagnosed during the trial.
Significance
Previous large, randomized trials using a combination of CA125 and transvaginal ultrasound have demonstrated similar positive predictive values (24–27%) [6,7] to that observed in this study; however, the cancer detection levels were low (40–60%) and this aspect, as well as the cost, has been the major drawback in the acceptance of this combined test as a screening test. The introduction of the CA125 rate-of-change algorithm, as used in the MMS design, has led to an increase in sensitivity from 60–86% [4] or 89.4% in this study [1]. Although this is clearly a better result, there are still issues with overall sensitivity. In recent transvaginal ultrasound studies [8] only 70% of postmenopausal women were detected with at least one ovary owing to limitations of ultrasound detection of the postmenopausal ovary. While CA125 is excellent for detecting serous, endometrioid and clear-cell epithelial cancers (sensitivity >90%), it is less suitable for mucinous tumors (70–80%) and ovarian sex cord cancers, such as granulosa cell tumors (30%), at least as reported with single CA125 estimates [9]. The corresponding values using the CA125 algorithm are unclear. It should also be noted that this study is confined to postmenopausal women who represent 79% [10] of women with primary invasive ovarian cancer. It is unclear if the MMS methodologies are appropriate in women of a reproductive age. These concerns can be addressed with the introduction of additional biomarkers that can help to identify the less common types of ovarian cancer.
In the trial, Menon et al. describe their findings as promising, recognizing that the study is not complete since the overall end point of the study is an assessment of deaths attributed to ovarian cancer and the ability of the screening tests to reduce the death rate. This aspect of the study will not be completed before 2014. Certainly, this interim report goes a long way towards the validation of an effective early detection test.
Executive summary
The combination of CA125 using a rate-of-change algorithm followed by transvaginal ultrasound as a second-line test provides a test with the requisite specificity to be considered as a screening test for ovarian cancer in postmenopausal women. Ultrasound alone was less effective.
The combined test, based on its high specificity (98.8%), resulted in 2.3 surgical operations per case of ovarian cancer, which is viewed as a satisfactory outcome and an improvement over other screening tests.
The sensitivity of the test (89.4%), while satisfactory, requires further validation. One approach would be to utilize additional complementary biomarkers.
This is an interim report. The trial's objective to establish whether these screening tests will reduce the death rate from ovarian cancer has yet to be assessed. This trial is expected to be completed in 2014.
Footnotes
This work was supported by a grant from the NHMRC of Australia. Robertson has an interest in a patented technology that has applications as an early detection test for ovarian cancer. The author has no other 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 apart from those disclosed.
No writing assistance was utilized in the production of this manuscript.