For years, Dr. Goff and colleagues have carried the banner for getting physicians to recognize that early stage ovarian cancer causes symptoms for months before diagnosis. For this they deserve accolades from gynecologists and primary care physicians, and appreciation from all the women we serve.
In this issue of Cancer, Goff et al report the results from a survey that was mailed to 3200 primary care physicians (gynecologists, family physicians, and internists).1 The respondents declared which tests they would order to evaluate patients who presented with various symptoms. The outcome measure was the proportion of physicians ordering the tests that would detect ovarian cancer: ultrasound, pelvic computed tomography (CT), and cancer antigen 125 (CA 125). The data essentially demonstrate that these physician groups have gotten the message: “Do not forget that early ovarian cancer causes symptoms.” This is a remarkable change in dogma. Before this knowledge, how many times had we explained to our patients with advanced ovarian cancer that we could not have made the diagnosis earlier; that the cancer already had metastasized by the time the patient had reported abdominal distension, bloating, and early satiety?
Some remarkable data emerge from this study. Nearly 90% of physicians reported that they would recommend at least 1 of the appropriate diagnostic tests; 71% chose an ultrasound, which, in my opinion, is the test most likely to indicate the risk of malignancy. This high rate of appropriate clinical management reflects positively on the efforts of Dr. Goff and others who have published on the symptoms of early stage ovarian cancer. Developing an educational process to improve this outcome would be very difficult.
Another encouraging finding was that gynecologists were more likely than other primary care groups to order the appropriate tests. The authors adequately explain that this is because of the core education of gynecology, which includes ovarian malignancies. Nevertheless, the appropriate response rates for family physicians and internists still were outstanding: 88% for both.
Less impressive findings of this study are the statistically significant but clinically insignificant outcomes. The large number of respondents enabled small differences to be statistically significant. The “adjusted” risk ratios of 1.07 (urologic [GU] vs gastrointestinal [GI] symptoms), 1.06 (group practice vs solo practitioner), and 1.04 (clinical teaching, yes or no) were all statistically significant, but a 4% to 7% relative difference in these variables does not seem to have much practical clinical value. For example, both GI symptoms and GU symptoms triggered excellent responses for ordering appropriate tests, 86% and 92%, respectively, a statistically significant difference. In my opinion, this 6% difference is relatively minor and should not overshadow the positive tribute that the vast majority of physicians would make the diagnosis of ovarian cancer in women presenting with either GI or GU symptoms.
Dr. Goff and colleagues report demonstrates that primary care physicians have made progress in the drive to diagnose ovarian cancer at an earlier stage to reduce the overall mortality of the disease. To date, a major need for this drive has been an adequate screening method. These authors and others have pointed out that the low prevalence of ovarian cancer, which is 1 per 2500 women aged >50 years, requires a screening method that has nearly perfect specificity, 99.6%, to provide a positive predictive value of only 10%.2, 3 Extremely high specificity is necessary to prevent the morbidity of invasive surgical procedures on patients with false-positive screening tests. Ironically, this challenge also applies to diagnosing early disease by symptomatology. Obviously, these abdominal, GI, and GU symptoms are nonspecific, and the prevalence in control populations varies greatly. Although roughly 10% of healthy women who present for annual gynecologic examinations report these symptoms, >70% of general primary care patients have these recurring symptoms.4, 5 This high prevalence of common symptoms lowers the specificity of pelvic imaging and raises the question of society's ability to pay for the number of tests necessary to diagnose 1 case of ovarian cancer. Despite this admonition, the cost of testing symptomatic patients intuitively would be less than screening all asymptomatic patients.
Bimodality screening, in fact, has reached the specificity that experts have estimated is necessary for an acceptable positive predictive value. Recently, the United Kingdom Collaborative Trial of Ovarian Cancer Screening published its results from screening over 200,000 postmenopausal women.6 That randomized control trial compared annual CA 125 plus transvaginal ultrasound (TVUS) as a second-line test (multimodal), TVUS, and controls. The sensitivity, specificity, and positive predictive values were 89.5%, 99.8%, and 35.1%, respectively, for the multimodal screen and 75%, 98.2%, and 2.8%, respectively, for TVUS alone. The number of surgeries to diagnose 1 cancer was 2.8 in the multimodal group and 34 in the TVUS group. Thus, the multimodal prevalence screen had specificity high enough to decrease unnecessary surgeries to a reasonable number. In fact, this ratio of surgeries per detected cancer may be the best reported to date in the literature, which has ranged from 3.7 to 163.2 In addition, the proportion of early stage cancers diagnosed by either method was approximately 47%.
One of the realities of all screening trials has been the discovery of advanced-stage disease mixed with the proportion of early stages. Investigators and practitioners may have to modify the goal of screening to be the detection of low-volume disease rather than stage I disease. Although the 5 year disease-free survival rate of patients with early stage cancer ranges between 80% and 90%, the median survival of patients with advanced disease that is completely resected is 5 to 6 years. Thus, any screening modality that enables the diagnosis of low-volume disease, regardless of stage, will dramatically impact ovarian cancer mortality in this country.
The objective of detecting early stage disease by either a screening methodology or clinical evaluation of early symptoms presupposes that ovarian cancer progresses in an orderly fashion from preinvasive disease, to early stage, to advanced stage. Let us assume that this is true. Is the preinvasive disease present long enough for screening methods to detect it in this form? Sophisticated modeling techniques applied to pathology results from prophylactic removal of tubes and ovaries in breast cancer 1 (BRCA1) gene mutation carriers provide educated propositions for serous cancers.7 The generated model suggests that these serous cancers persist for more than 4 years in the in situ, stage I, or stage II phase and for approximately 1 year as stage III or IV cancers before they become clinically significant. However, disappointingly, the model also predicts that these cancers measure <1 cm for most of the occult period and that screening methods would have to detect the tumors at 1.3 cm to prevent 50% of them advancing to stage III. These small tumor sizes usually are below the detection threshold of imaging techniques.
Recently, evidence has been accumulating to challenge the supposition that all ovarian cancers progress in an orderly fashion. Indeed, molecular markers have identified 2 types of epithelial ovarian cancer.8-10 One group is composed of low-grade serous, low-grade endometrioid, clear cell, and mucinous tumors. These are indolent tumors and are more likely to be diagnosed in stage I. They are genetically stable and have mutations in v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS); v-raf murine sarcoma viral oncogene homolog B1 (BRAF); phosphatase and tensin homolog (PTEN); and transforming growth factor, beta receptor II (TGFbR2). These tumors may exhibit a morphologic continuum from adenoma, to borderline malignant, to malignant. The second group of epithelial tumors, type II, includes the high-grade serous, high-grade endometrioid, and undifferentiated tumors, which are known for their aggressive behavior and diagnosis in advanced stage. Over 80% of these type II epithelial tumors display P53 mutations, some are related to BRCA dysfunction, and they rarely contain the mutations associated with type I tumors. These tumors are genetically highly unstable. Thus, screening tests and symptom identification are likely to detect type I tumors in an early stage and type II, aggressive serous cancers in an advanced stage. However, if these latter cancers are detected early when the tumor volume is low, with a likelihood to be completely resected, then the goal of lowering mortality by screening is realistically achievable.
One additional recent discovery that impacts early diagnosis of epithelial ovarian cancer is the identification of the fallopian tube mucosa as the probable source of many of the serous cancers, both primary ovarian and primary peritoneal.11-14 Serous tubal intraepithelial carcinoma (TIC) has been identified in the majority of these high-grade serous carcinomas, and invasive serous cancer is present in the same tube in 70% of these patietns.12 If aggressive serous carcinomas arise from preclinical TIC, then screening methodology needs to shift emphasis from detecting adnexal enlargement, with its possible associated symptoms, to targeting TIC. Such a strategy would need to include molecular markers associated with TIC.
Tremendous resources are being consumed to identify “tumor markers” for ovarian cancer. Greater than 30 serum markers have been evaluated alone and in combination with CA 125.3 Proteomic technology is advancing and promises to improve screening results.1 Now that serous TIC has been identified as the potential “precursor” of the most lethal ovarian cancers, researchers will have the challenge to identify a very specific tumor biomarker that can detect the cancer in its preclinical phase. This may require novel approaches beyond the traditional serum protein biomarkers.7 Because the fallopian tube and the endometrium (for high-grade endometrioid cancers14) are suspected sites for precursor lesions, sampling these areas for biomarker(s), once identified, may be the screening procedure of the future. Cervical cytology and endometrial cytology already have sensitivities for diagnosing all stages of ovarian cancer up to 19% and 42%, respectively.15 And, just perhaps, future biomarkers could increase the sensitivity of early diagnosis by these means. Maybe someday we will have a “Pap smear” for ovarian cancer.