The search for meaning—Symptoms and transvaginal sonography screening for ovarian cancer

Silent no more

Ovarian cancer remains the leading cause of death from gynecologic malignancy among women in the United States.1 Much of the mortality of ovarian cancer is attributable to the finding that most women present with late-stage disease, when cure rates are low. Although survival rates for patients with early stage disease approach 80% to 90%, only a minority of patients will be diagnosed with localized disease.2 Efforts to identify effective screening strategies have been disappointing: There has been little actual shift in the stage at which screened patients are diagnosed with cancer, and the positive predictive values are unacceptably low. Consequently, ovarian cancer screening currently is not advised for the general population.3, 4 Lacking a reliable screening tool, some clinicians have focused attention on the earlier recognition of symptoms associated with ovarian cancer to make the diagnosis earlier. Aggregate data suggest that patients with ovarian cancer do have symptoms that precede their diagnosis by months.5-7 Recent data have identified a constellation of specific symptoms (a symptoms index) that may predict ovarian cancer and requires validation in larger patient populations.8-10

In this issue of Cancer, Pavlik et al propose to test how this symptoms index performs compared with transvaginal ultrasound in the detection of ovarian cancer.11 To answer the question, the authors draw on their published experience with the University of Kentucky Ovarian Cancer Screening Project spanning from 1987 to 2008 that has performed transvaginal ultrasound on over 31,000 women.12 Four hundred fifty women from this cohort required surgery for an abnormal ultrasound using a morphology index schema to incorporate ovarian volume and/or abnormal sonographic features, which included both benign and malignant tumors. The authors then approached patients retrospectively with a 1-time symptoms questionnaire meant to reflect any symptoms that they had at the time of their suspected pathology before surgery. The study group of 272 patients responded to the questionnaire, which was derived from a symptoms survey developed by Goff et al to predict ovarian cancer. In this model, a symptoms index was considered positive if a patient experienced any of 6 symptoms (pelvic/abdominal pain, increased abdominal size, bloating and feeling full/difficulty eating) for <1 year that occurred >12 times per month.8, 9

Pavlik et al observed that 27 women (10%) reported symptoms that were considered positive on the symptoms index, whereas an equivalent proportion of women had no symptoms or symptoms that were not considered informative. Only 6 of the 30 women with ovarian cancer had a positive symptoms index (20% sensitivity), although all the true positives had early stage disease. Ultrasound performed considerably better as a screening tool than the symptoms index, and no combination of symptoms index and ultrasound findings significantly improved the detection of ovarian cancer compared with ultrasound alone. Ultimately, the only value of combining both modalities was the enhanced specificity that resulted when both tests were negative. In addition, the authors suggest that the clinical management of a minimally abnormal ultrasound should factor in patient symptoms to reduce surgery on benign masses.

The disappointing performance of the ovarian symptoms index in the current study, compared with other reports, is not unexpected given the differences in study populations and study design.5-9 The high frequency of reported symptoms in women with pelvic masses or predominantly advanced-stage ovarian cancer was not observed in the current study of asymptomatic women in a large screening program with lower disease prevalence. Unlike previous studies, the majority of women with ovarian cancer in this study had early stage disease. Although the 6 patients who had informative symptoms all had early stage disease, the remaining 17 patients with early stage disease did not have a positive symptoms index. It is possible that women with early stage disease may have a different cluster of symptoms than women with advanced-stage disease and may require an expanded symptoms inventory in future studies. It is noteworthy that Andersen et al reported that the addition of the CA 125 tumor marker to the Goff symptoms index enhanced the detection of ovarian cancer in a case-control study, most notably in early stage disease, with a sensitivity of 80%.10

Unfortunately, the current study has the same design flaw of many studies that have relied on the retrospective recall of women who required surgery.13 To minimize the recall bias for the potentially lengthy interval between the time of surgery and the patient's completion of the symptom questionnaire, in some cases, the authors introduced a confidence assessment question to allow for the discrimination of uncertainty as exclusionary criteria. The rating of this confidence response is presented with little justification or validation and begs the question of how patient confidence can be scored reliably. The exclusion of 13 women in the study who had confidence response scores <3 appears somewhat arbitrary. Despite the authors' attempts to compensate for this potential bias, they cannot answer their own study question with the current study design. Prospective collection of symptoms information from women who are followed in a prospective screening program is the more appropriate way to study the potential benefit of both screening tools. The authors' conclusions regarding the minimal incremental gains they observed should not dissuade others from further study of this important research question.

Another concern is the exclusion of the other 165 women (36%) from the original 450 study participants who required surgery but did not want to participate, had died, or could not be contacted. Although the authors demonstrate similar clinical and surgicopathologic characteristics between the original 450 women and the study group of 272 women, this exclusion poses some limitations given a potential relation between the women who refused to or were unable to complete the symptom questionnaire and their responses. In view of the poor survival of ovarian cancer, the prolonged time interval between diagnosis and completion of the survey introduces another inherent patient bias.9

Pavlik et al note that the strength of their study design is that they draw their patients from a large cross-section of the population that was followed prospectively and that all patients required surgery. Therefore, the control group is more representative of the general population and provides a more relevant comparison between women who have and do not have symptoms. In contrast, the control groups in other studies have included a heterogeneous group of women ranging from women seeking no care, to those referred for ultrasound, to high-risk women in an ovarian cancer early detection study.5-10

Regardless of how and which symptoms index is included in future ovarian cancer screening programs, the true value of the findings by Pavlik et al is the discussion this article will generate regarding ovarian cancer symptoms. Goff et al have clearly established that these symptoms cannot be ignored and that women with informative symptoms do have an increased risk of having ovarian cancer.5, 8-10 A recent randomized pilot study in a population of women aged >45 years confirmed the feasibility of screening for ovarian cancer using symptoms associated with ovarian cancer. A study group of general practitioners was educated about the typical symptoms associated with ovarian cancer and was given the opportunity to refer patients for targeted screening using CA 125 and ultrasound. Specificity was high, patient compliance rates were good, and there were low rates of intervention and additional investigations.14

Patients with ovarian cancer have taught their physicians that ovarian cancer is not the “silent killer” presented in our medical textbooks. This process of patient and physician communication will drive the research to better define early symptoms of ovarian cancer and spread the word about symptom recognition. Several studies have described that the majority of patients with advanced ovarian cancer experience symptoms for several months before the appropriate tests and referrals establish a diagnosis. Improved physician education and patient awareness regarding ovarian cancer symptoms may minimize this delay. Although the clinical significance of this brief time interval is unknown, it is plausible that establishing a diagnosis even months earlier may reduce the patient's tumor burden, which has been correlated directly with the likelihood of optimal surgical cytoreduction and patient survival.2, 9, 15

Until an effective screening test for ovarian cancer is found, the clinical challenge remains how to discriminate between significant and common symptoms to best care for patients. Given the vague nature of these symptoms that lack an exact threshold for further costly evaluation, this is no small task. The optimal constellation of symptoms to screen for ovarian cancer may never be defined; however, progress has been made to recognize that such symptoms do exist. The silence has been broken.

Conflict of Interest Disclosures

The author made no disclosures.