The trusted source for
healthcare information and
Will Ovarian Cancer Screening Really Impact Mortality?
Abstract & Commentary
By Robert L. Coleman, MD, Professor, University of Texas; M.D. Anderson Cancer Center, Houston, is Associate Editor for OB/GYN Clinical Alert.
Dr. Coleman reports no financial relationship to this field of study.
Synopsis: Despite the marked survival differences between early and advanced stage ovarian cancer and the promise of "stage migration" with general population screening, modeling suggests the impact will be only modest due to identification of low-risk disease.
Source: Havrilesky, et al., Development of an ovarian cancer screening decision model that incorporates disease heterogeneity. Cancer 2011;117:545-553.
Recent investigation into the molecular pathogenesis of epithelial ovarian cancer has implicated two dominant phenotypes. One manifests by late presentation, advanced stage, and an aggressive clinical course (Type I), and one with a more indolent nature, which, despite an innate chemo-resistance, is associated with long survival (Type II). By studying the natural history of ovarian cancer (Type I vs Type II), the authors sought to evaluate the impact of screening on mortality. They considered both a 1-phenotype and a 2-phenotype model, the latter including the assumptions on outcome based on the contribution of Type II cancers. To calibrate their data, they used stage distribution and survivorship data from the National Cancer Institute's Surveillance Epidemiology and End Results (SEER) database. They also assumed their "screening model" would perform in line with the multimodal screening algorithm (MMS) used in the recently reported U.K. Collaborative Trial of Ovarian Cancer Screening (UKCTOCS) study, and they adjusted the SEER prevalence ovarian cancer rate to match that reported in the UKCTOCS study. The authors showed that their validation test of screening performance would increase the number of stage I/II cancers (to about 41%), in line with that reported from the UKCTOCS study. Positive predictive value also was close (26%-27% vs 35%). Overall survival for ovarian cancer predicted based on the 1-phenotype and 2-phenotype model was very similar to that expected from the SEER data. Surprisingly, the impact on mortality from an implemented postmenopausal annual screening program resulted in only an 11% (2-phenotype model) to 15% (1-phenotype model) reduction in mortality. Modeling different screening characteristics (sensitivity and specificity) and frequencies adjusted these measures only slightly, with the exception of the screening frequencies, which had its greatest impact with every-3-month evaluation. The authors conclude that screening, such as that used in the UKCTOCS study, will likely have only a modest impact on survival, due largely to the more efficient discovery of indolent malignancy.
This is a provocative report and tempers some of the enthusiasm that surrounds the anticipated results (2014) from the impressive effort being conducted in the United Kingdom. Initial data from their prevalence study has raised hopes that stage migration of high-grade ovarian cancer would translate into notable survival gains. The current study draws attention to a basic tenet of successful and effective cancer screening: An identifiable pre-invasive or early invasive state, which, if identified, is associated with improved outcome. What the study suggests is that the more indolent cancer (which is associated with a better survival) will be more often found through annual screening largely due to its growth rate. However, the proportion of diagnosed indolent cancer is small relative to typical high-grade cancer. While a 10%-15% reduction in mortality is modest, it would be welcomed in any clinical trial evaluating a pharmaceutical agent, especially in recurrent disease.
While modeling in this fashion is helpful to address potential impact on population dynamics, it is limited by our assumptions. This study is no exception. Many pieces of data needed to be included and the sources for those data, as well as the magnitude for each of the variables, are quite subjective and could be inaccurate. For instance, problems with the SEER registry are notorious because of lack of validation and central review. However, potentially even more problematic was the author's inclusion of clear cell and mucinous cancers into the indolent phenotype. Our understanding of the biology driving these tumors, as well as low grade serous tumors, is quite different than high-grade serous; however, expected mortality from advanced stage clear cell and mucinous tumors are significantly higher than their "aggressive phenotype" and even more so relative to the low-grade serous tumors. Adjusting for anticipated stage at presentation made this cohort perform better in the analysis. Ultimately, the debate will be settled with the UKCTOCS survival analysis due in 2014. At that time we'll be able to determine whether ovarian cancer screening is wishful thinking or a new reality.