The most award winning
healthcare information source.
TRUSTED FOR FOUR DECADES.
Is There Value in CA125 Distinguishing Benign from Malignant Ovarian Tumors?
By Robert L. Coleman, MD, Associate Professor, University of Texas; M.D. Anderson Cancer Center, Houston, is Associate Editor for OB/GYN Clinical Alert.
CA-125 measurements alone or as part of a multimodal strategy are routinely used clinically to aid in developing a treatment plan for women with adnexal masses. Well-known limitations in CA-125 testing, particularly in premenopausal women, generally indicate additional measures—usually imaging, to most accurately describe the clinical state of affairs and to develop a therapeutic strategy. Timmerman and colleagues1 reviewed the performance of CA-125 alone and in combination with a number of mathematical models to predict the histologic nature of persistent adnexal masses in 809 recruited women participating in the International Ovarian Tumor Analysis (IOTA) study. Prevalence of cancer in this study was 30%, with approximately two-thirds representing primary invasive ovarian cancers; 55% of the women participating were premenopausal. Test performance was measured by calculating the area under the receiver operator characteristics curve for each of the models. For menopausal and premenopausal cohorts, a training set and testing set were used. The prevalence of cancer in the prospective test set was 37%. Overall, CA-125 did not add to the testing performance of the ultrasound-dominant malignancy index model (M1). The original M1 was constructed from 12 independent variables including personal history parameters. When new models were constructed in the pre- (designated "M3") and menopausal (designated "M4") cohorts, CA-125, once again, did not improve the testing performance of the models; however, CA-125 as a single parameter performed well relative to the model in menopausal women. The authors conclude that adding CA-125 to clinical information and ultrasound information does not improve the discrimination of mathematical models between benign and malignant adnexal masses.
Over the more than two and a half decades since CA-125 was discovered, the biomarker has proved to be profoundly useful in a number of clinically important scenarios, including prognosticating outcomes to therapy, monitoring response to adjuvant treatment, surveillance, and defining progression of disease. Its reliability in these arenas has also led to the incorporation of CA-125 as a measure of disease progression in clinical studies and is being considered as a measure of response by the Food and Drug Administration. It is also being evaluated prospectively in serial measures as a determinant of ovarian cancer development (screening) and has served in a surveillance role in primary malignancies and diseases other than ovarian cancer. On the whole this single biomarker is incredibly useful. However, it is not perfect —and, as demonstrated in the current study, CA-125, as a one-time measure, may not be the most accurate way to discriminate malignancy in women with adnexal masses.
The authors of this report have championed the utility of clinical history, such as history of ovarian cancer, hormone use, and age combined with specific ultrasound characteristics as the most efficacious way to identify a probability estimate of malignancy. The impetus to develop and validate a model such as this is to help with the triage of patients to the appropriate specialists (gynecologic oncologists) should the index of suspicion for cancer be high. The study was well-conceived and conducted and used a prospective testing set with rigorous parameters to compare the various models. Although the sample is larger than most others reported to date, the number of cancers in the subgroups is still relatively small upon which to make inference. In addition, the prevalence of cancer among the cohorts by history was likely not equivalent nor large enough to evaluate a differential effect. For example, the probability of malignancy in a premenopausal woman with an adnexal mass in this study was much lower than the likelihood of malignancy in a woman with a history of ovarian cancer and an adnexal mass. Nonetheless, there are other considerations to note, which limit the "common" utility of the report's conclusions.
First, while the model is statistically derived, it has detailed information about the patient that is not always available to the sonographer. The regression formula contains 7 to 12 variables (depending on menopausal status) and cannot be practically calculated without a computer. Such a limitation can be overcome but it raises the next question: "what's a significant probability estimate in which to become concerned?" Is any number about 10% important enough for referral? For surgery? For repeat testing? In addition, highly trained sonographers retrieving specific information on the mass(es) led to the model's construction. Are such characteristics reproducible outside of a tertiary care referral center? Last, while CA-125 was dropped in these models as not adding significantly to the testing performance, several surrogates likely overshadowed its impact, such as ascites and mass size—particularly the solid component.
The clinical utility of biomarkers is immense and subject of intense investigation. The advantage a blood test as over operator-dependent testing is obvious in its generalization. However, the search for equally performing biomarkers has been problematic, though progress is being made. It is likely that future reports will include a panel "cocktail" in which to develop a risk index. However, decision algorithms based on these results will need validation for clinical practice.2-3
Any effort improving the precision of estimation is worthwhile. The trick will be in developing a clinically useful model that is applicable to the population in general.