Clinical Performance Comparison of Two In-Vitro Diagnostic Multivariate Index Assays (IVDMIAs) for Presurgical Assessment for Ovarian Cancer Risk

The detection of a pelvic adnexal mass increases the risk for pelvic malignancy and thus initiates an evaluation to determine optimal management [1]. An adnexal mass can be the result of a gynecologic or nongynecologic process; however, the approach taken to determine the malignant potential of this finding is related to a complex and multivariate set of demographic and clinical findings including, but not limited to, the age of the patient, the appearance of the mass on imaging, personal and family histories and symptom presentation [1]. The evaluation of a woman with an adnexal mass will invariably lead to a variety of potential interventions, from monitoring without medical or surgical intervention to pharmacologic intervention or surgical extirpation. Of importance is that in cases of adnexal masses associated with pelvic malignancy, optimal outcomes are obtained when the initial surgical procedure is performed by a surgeon skilled and experienced in providing effective surgical management and tumor staging within a facility experienced in caring for such patients [2].

The development of an assay or algorithm that would provide a more accurate and simple approach to malignancy risk assessment in cases of pelvic adnexal masses has been a long-sought-after goal. Such an assay or algorithm would potentially provide information to allow for the appropriate triage so as to improve clinical outcomes without overwhelming certain clinical services, especially those providing surgical oncologic care given the more than 300,000 women who present with an adnexal mass in the USA each year [3, 4]. Several algorithms have been developed that use a variety of serum biomarkers, patient demographics and imaging. The FDA has cleared three IVDMIAs that employ measurements of serum biomarkers for the evaluation of women presenting with a pelvic mass: MIA or Ova1^®, a second-generation multivariate index assay (MIA2G, Overa) and the Risk of Ovarian Malignancy Assay (ROMA). These assays provide a relatively simple and straightforward approach to assessing the malignant potential of an adnexal mass. In this article we present the findings of a comparison study of ROMA to the most recently developed second-generation MIA (MIA2G).

Subjects were enrolled prospectively at 44 sites across the US from primary care women’s health clinics, obstetrics and gynecology groups, gynecologic oncology practices, community and university hospitals, and health maintenance organizations. The subjects were drawn from two previously published trials, OVA1 and OVA500 [5, 6]. Both trials used the same inclusion and exclusion criteria, with inclusion criteria being females 18 years of age and older, documented ovarian tumor with surgery planned within 3 months of imaging, consent to phlebotomy and a signed informed consent. Exclusion criteria were women younger than 18 years of age, a previously diagnosed malignancy in the previous 5 years (with the exception of non-melanoma skin cancer), no planned surgical intervention or lack of consent for a surgical procedure or phlebotomy. For both studies, menopause was defined as lack of menstruation of 12 consecutive months or age 50 or older when menstrual history was not noted. This study was performed in accordance with the Helsinki Declaration of 1964 and its later amendments; institutional review board approval was obtained from each individual site (Table S1).

Participant data from two previously performed and published prospective clinical trials were combined (OVA1: February 2007–April 2008; OVA500: August 2010–December 2011) to form the overall study cohort. Of the 1110 women enrolled in the two previous trials, 993 subjects were evaluable: 500 from the OVA1 study and 493 from the OVA500 study (Fig. 1). Reasons for excluding study subjects from analysis included: surgery was either not performed or delayed > 3 months, pathology report was not available, blood specimen was unusable, physician assessment was not available, and imaging study did not confirm an adnexal tumor. Table 1 displays the breakdown of subjects used in the study by pathology and menopausal status. The distribution of subjects between menopausal states was approximately equal, being 51% premenopausal and 49% postmenopausal.

Of the total population, 245 (24.7%) were diagnosed with a malignancy. The proportion of malignancies among the postmenopausal subgroup was approximately twice the rate for the premenopausal group, which is consistent with menopausal status being a known risk factor for gynecologic cancers. Of the malignancies detected, 166 (67.8% of the malignancies) were primary ovarian malignancies; 42 (17.1%) were (borderline) low malignant potential tumors; 23 (9.4%) were non-primary ovarian malignancies with involvement of the ovaries, and 14 (5.7%) were non-primary ovarian malignancies with no ovarian involvement.

The performance of biomarkers and biomarker panels was assessed in the absence of other physician assessment tools. As shown in Table 2, the evaluation of these subjects’ serum samples by the three assays demonstrated sensitivities of 71.0% (CI 65.0–76.3) for CA125, 79.2% (CI 73.7–83.8) for ROMA, and 91.0% (CI 86.8–94.0) for MIA2G. MIA2G’s sensitivity is statistically significantly higher than that of ROMA, as calculated by McNemar’s chi-square test. Specificities for the three assays were 89.7% (CI 87.3–91.7) for CA125, 78.9% (CI 75.8–81.7) for ROMA and 65.5% (CI 62.0–68.8) for MIA2G. CA125 had the lowest rate of false positives at 7.8% overall. MIA2G showed lower specificity than either ROMA or CA125 with a false-positive rate of 26.0%. Overall, MIA2G had the lowest rate of false negatives, 2.2% overall and 22 out of 245 total malignancies (Table 3). CA125 had a false-negative rate of 7.2% overall, or 71 out of 245 total malignancies. ROMA fell in between with a false-negative rate of 5.1% overall, or 51 out of 245 total malignancies.

The clinical performance of the assays stratified by the subjects’ menopausal status is presented in Table 4. CA 125 performance was based on defined menopausal status-dependent cutoff values [5]. Premenopausal sensitivity for CA125 was 50.7%. ROMA and MIA2G also showed somewhat higher sensitivity for postmenopausal subjects, but the difference was not as great as that for CA125. Regarding specificity, both MIA2G and CA125 demonstrated higher values among premenopausal subjects, whereas ROMA was somewhat higher for postmenopausal. Overall, the lowest false-positive rate was that for premenopausal CA125 at 3.6%, and the lowest false-negative rate was premenopausal MIA2G at 1.4%. As with the overall sensitivities and specificities, ROMA generally fell somewhere between the two.

Table 5 shows the sensitivities of the three assays characterized by tumor histology. As seen in the overall clinical performance, CA125 had the lowest sensitivity, MIA2G had the highest, and ROMA fell in between. This was true for all the histologic subtypes of malignancies detected in this study: primary ovarian (epithelial and non-epithelial), low malignant potential (LMP), metastatic with ovarian involvement and non-ovarian. All three assays had the highest sensitivity for primary epithelial ovarian malignancies: 83.3% (CI 76.6–88.4) for CA125, 89.3% (CI 83.4–93.3) for ROMA and 96.7% (CI 92.4–98.6) for MIA2G. The lowest histologic subgroup sensitivities were primary non-epithelial ovarian malignancies for CA125, which was 37.5% (CI 18.5–61.4), and for MIA2G at 75.0% (CI 50.5–89.8). ROMA’s lowest was for LMP masses, with a sensitivity of 50.0% (CI 35.5–64.5).

The sensitivities are also characterized by stage and grade of the primary ovarian malignancy in Table 6. One considerable challenge related to the relative dearth of successful outcomes in ovarian cancer is the relatively high frequency of advanced stage disease at the time of detection, resulting in higher rates of adverse outcomes compared with tumors detected at a less advanced stage [7]. However, in this study a higher rate of early detection was evident, possibly leading to longer life expectancy that may be further facilitated by recent findings showing correlation between specific deleterious somatic genomic variants and improved outcomes with certain chemotherapeutic regimens [8]. The combined early stage (I and II) sensitivities were 63.1% (CI 52.4–72.6) for CA125, 76.2% (CI 66.1–84.0) for ROMA and 90.5% (CI 82.3–95.1) for MIA2G, as opposed to the sensitivities for late stage cancers, which were 97.5% (CI 91.3–99.3) for CA125 and ROMA and 100% (CI 95.4–100.0) for MIA2G. MIA2G is also notable for correctly identifying all stage III cancers, which both ROMA and CA125 were unable to accomplish. A similar phenomenon could be seen for cancer grade, with grade 3 malignancies having the highest sensitivities across all three assays for those that were graded followed by grade 2, with grade 1 having the lowest sensitivities. For all stages and grades, the generally observed trend of MIA2G having the highest sensitivity, followed by ROMA and then CA125, is also present. It is also noteworthy that MIA2G had a 100% sensitivity for both late stage and grade 3 (serous) primary ovarian malignancies.

The detection of an adnexal mass, regardless of the manner in which it was detected, should prompt an evaluation to determine the need for intervention versus the ability to monitor without intervention [1]. This decision is one fraught with complexity, ranging from the presence of symptoms at the time of detection to the age of the patient, including her family and personal histories as well as the appearance of the mass with imaging modalities [1]. As such, there is no single or simple approach to determine the most appropriate management of the adnexal mass. Given the increased risk for malignancy in these types of cases, this assessment should take into account all factors that best serve to characterize the benign or malignant potentials of the clinical finding. Urban and colleagues [9] recently showed that combining an assessment of symptoms with MIA2G to evaluate women presenting with an adnexal mass further improved the accuracy of the triage assessment for ovarian cancer.

It is obvious that the detection of an adnexal mass requires an accurate assessment of its malignant potential to best provide ongoing care. This assessment is even more critical given that most women found to have an adnexal mass do not have immediate access to a gynecologic oncologist or surgical oncologist who practices within a hospital-based system experienced in caring for women and men with cancer. The inability to accurately assess the malignant potential of an adnexal mass thus creates a profound clinical dilemma: either refer all women with an adnexal mass to a center with trained cancer professionals and overwhelm that system with many women who have benign adnexal lesions or surgically evaluate many if not most such women and then refer only those women with pathology-proven malignancies for care by a cancer center. Indeed, an often-practiced approach by some non-oncologic surgeons and gynecologists has been to perform a surgical procedure to obtain a tissue sample solely for malignancy confirmation. Once cancer is confirmed by pathology, the patient is then sent to a cancer center for further treatment and care. However, this practice serves to reduce the longevity and quality of life for that woman, given the fact that optimal clinical outcomes are achieved when women are cared for by skilled cancer physicians in experienced cancer care facilities at the time of their initial procedures [2]. Regardless of the economic impact of referring a surgical patient elsewhere, it behooves every clinician to make every attempt to gain as much insight as possible into the adnexal mass prior to intervention so as to optimize clinical outcome. In cases of adnexal mass, the ready availability of triage algorithms such as ROMA and M1A2G could considerably reduce the frequency of needless and harmful surgical procedures for malignancy confirmation that adversely impact the patient and her prognosis.

In this study, we found that all triage approaches provided invaluable information that could serve to provide the best care to women with an adnexal mass. For those at a low risk for malignancy, surgery close to home by a skilled gynecologic surgeon would provide good care without the need to travel long distances. In this comparative study, an assessment of the overall clinical performance showed CA125 to have the lowest sensitivity and MIA2G to have the highest. This was true for all histologic subtypes of malignancies detected in this study: primary ovarian (epithelial and non-epithelial), low malignant potential (LMP), metastatic with ovarian involvement and non-ovarian. Conversely, MIA2G showed lower specificity than either ROMA or CA125, though MIA2G had the lowest rate of false negatives.

Despite our findings, as well as those of other studies that characterize CA125 to be considerably less sensitive and specific than multimodal risk algorithms such as MIA2G or even ultrasound [10] for determining the likelihood of malignancy, the use of CA125 alone to assess women with adnexal findings remains a popular choice in the assessment of women presenting with an adnexal mass. A recent study from Italy by Piovano and colleagues [11] found that adding CA125 and ROMA to triage algorithms for postmenopausal women with adnexal masses improved screening accuracy compared with ultrasound alone; the addition of ROMA was considerably more expensive and provided for minimal additional screening benefit compared with CA125 alone. While economic considerations do play some role in the continued use of CA125 alone for the presurgical assessment of women with adnexal masses, the availability of FDA-cleared multimodal assays and the lack of FDA clearance for the use of CA125 in these clinical situations continue to call into question the ongoing use of a single analyte assay for which there is considerable evidence to support it being an inferior triage process for women presenting with adnexal masses compared with multimodal algorithms [12]. In addition, it begs repeating that the use of any multimodal assay in such cases should not be used for the initial evaluation of an adnexal mass or for the determination of clinical management in such cases. Rather, an IVDMIA should be used solely to determine optimal personnel and venue should surgical intervention be chosen to manage the adnexal mass.

Our study is impacted by several limitations. First, while this study was a prospective enrollment study, a considerable number of participants did not complete the study requirements for the reasons listed above. It is unclear whether the exclusion of these enrollees had an impact on the outcomes of our study. Another limitation was the lack of a consistent definition of adnexal mass used throughout the study centers. This was done to create a more “real-world” study, but may have led to the inclusion of inappropriate subjects or the exclusion of appropriate subjects.

Our study shows that while all algorithmic approaches to adnexal mass triage provide valuable information concerning malignant potential, each algorithm has unique triage characteristics that will appeal to particular clinicians. M1A2G is characterized by the highest sensitivity and lowest false-negative rate in this comparison study, thus making it the test most likely to correctly characterize an adnexal malignancy with the lowest likelihood of a missed malignancy. However, this assay will also characterize more women with benign findings as having an increased risk of malignancy than the other two assays studied. Regardless of the triage algorithm used by an individual clinician, the use of a triage algorithm will invariably help guide clinicians in their decision to best manage a woman presenting with an adnexal mass, thus increasing the likelihood that the care provided to that woman will be appropriate and, in cases of pelvic cancer, maximize the survival rate by optimizing that woman’s initial surgical intervention and care.