Summary of Evidence

Since the previous review of this topic, mortality results from the large, well-designed UKCTOCS were published. Thus, there were three trials reporting mortality outcomes after ovarian cancer screening. The PLCO and UKCTOCS^{91, 92} were designed with statistical power to detect a 30 to 35 percent difference in mortality from this relatively rare but often fatal cancer, and had null findings in primary analyses. The U.K. Pilot trial⁹⁸ reported mortality outcomes, but was designed to examine the feasibility of screening and was underpowered to detect a mortality difference. The small QUEST trial⁹⁸ evaluated quality of life and psychological effects of ovarian cancer screening. Given evidence that there is no mortality benefit from ovarian cancer screening, the harms associated with screening merit focused consideration. For women in the ovarian cancer screening programs evaluated, positive tests and followup can lead to surgery and surgical complications in women without disease.

Further Discussion of UKCTOCS

Despite a primary null finding from the UKCTOCS on ovarian cancer mortality, the investigators included statistical analyses suggestive of a possible long-term benefit of the CA-125 ROCA screening intervention on ovarian cancer mortality (excluding peritoneal cancers), based on their observation that Kaplan-Meier cumulative mortality curves appear to diverge approximately 10 years after randomization. Consistent with best practices,⁹⁴ the trialists were transparent in their reporting of protocol-specified versus post hoc exploratory analyses. There are several reasons our review did not focus on these secondary analyses. First, we prioritized analyses with both ovarian and peritoneal cancer included, since their presentation and treatment is not distinct in clinical practice and because they are often difficult to distinguish pathologically. More of the cancers identified in the CA-125 ROCA screening arm of UKCTOCS were coded as peritoneal cancers than in the ultrasound and no screening arms. It is not surprising that excluding these cases that have very high mortality increases the difference in mortality between the CA-125 ROCA and control arm, Nevertheless, excluding peritoneal cancer cases did not alter the null findings of the pre-specified statistical analyses (Appendix D). Second, we focused on statistical tests that were specified a priori through publication of a protocol and trial registration.^{107, 108} A statistically significant benefit for the CA-125 ROCA intervention was also detected when peritoneal cancer cases were excluded and a weighted Log Rank test was used, which assigns greater weight in the survival analysis to later years of the trial (HR 0.78 [95% CI, 0.62 to 0.97]). This report focuses on the prespecified statistical analysis. Beyond the 10 year followup range of the trial, far fewer of the study participants have accrued followup data, and statistical analyses weighting the years far beyond the end of the screening period may overemphasize the observed later divergence of the ROCA arm. Nevertheless, publication of UKCTOCS trial data incorporating additional years of completed followup will be important contributions, as will causal theories to explain delayed screening effects, if observed.^107-109 A subgroup analysis of the data that obtained statistically significant findings endeavored to remove cases that were prevalent at the outset of the trial, using imputation, modeling and stored CA-125 data when available. Our review focused on intention-to-treat analysis of all participants, since these findings are most robust and applicable to the implementation of a screening program, and its cumulative effects.¹⁰⁹

Third, the divergence of the trial arms later in the study period are more difficult to attribute to the original randomized condition and screening, as the longer a study continues, the more opportunities there are for measured and unmeasured differences in the study arms to accrue. As noted, there were also substantially fewer women at risk included in the analyses beyond 10 years, because women recruited into the study later have not yet accrued followup for inclusion in the analysis of the longer time horizon. Thus, data from the later years of the trial (>10 years) are based on incomplete data and should be cautiously interpreted.

Differential reasons for censoring could lead to some divergence in the ovarian cancer mortality curves as followup times lengthen. In this study, there were no differences across arms in participant followup (censoring) or other causes of death, but there may have been differences between arms and changes over time in the proportion of participants in the trial with two ovaries intact. Those with both ovaries rather than just one, by definition, have higher ovarian cancer risk. In other words, the usual-care screening arm may have had a net surplus of ovaries at risk, despite a similar proportion of women at risk. We calculated the overall proportion of women having an oophorectomy during the trial, based on surgical investigation of screening results (true and false positives) or other indications based on numbers reported for the UKCTOCS. Approximately 4.4 percent of women in the ultrasound arm, 2.3 percent in the CA-125 ROCA arm, and 1.4 percent in the no-screening arm had oophorectomies (commonly bilateral, with or without salpingectomy). Others have suggested that the potential prophylactic effect of ovary and fallopian tube removal might influence the UKCTOCS results, especially in the long term.¹¹⁰ To date, there is no overall difference in the incidence of ovarian cancer by arm, suggesting that a prophylactic effect is not present, but as more years of followup data are available for more of the enrolled participants, additional analysis of the cumulative cancer incidence rate by study arm can be undertaken. The UKCTOCS team have received grant funding to continue follow up through 2018.¹¹¹

Further Discussion of PLCO

The PLCO was the only U.S.-based study for directly assessing potential net benefits or harms of screening. The trial protocol consisted of annual transvaginal ultrasound for 4 years and annual CA-125 testing for 6 years. Annual bimanual ovarian palpation by trained examiners was also included during the first 4 years of the trial.⁷⁰ Depending on when women entered the study, they received 0 to 4 physical palpation examinations as part of the screening program, consequently 20,872 women in the screening arm received at least one ovarian palpation examination (61%). This screening modality was dropped from the trial, however, because no cancer cases were identified solely on the basis of palpation (i.e., all cancer cases were also positive on CA-125 and/or TVU), and test sensitivity (defined as cancer diagnosed within 1 year of screening positive with a palpable adnexal mass) was very low (5.1%).

Overall, in the PLCO, the ovarian cancer mortality rate was greater in the intervention arm compared with usual care over 13-years of followup, although the difference was not statistically significant.⁹¹ A recently published analysis added up to 6 additional years of post-trial followup mortality data (mean 2.3 years) and did not find evidence of a late-emerging benefit of screening.¹¹² The post-trial data were obtained through a different, centralized system rather than through the trial screening centers, and upon completion of the trial centralized followup was refused by 16 percent of usual care women and 12 percent of intervention arm participants. Consequently, total followup time was shorter for refusers than for those willing to participate in ongoing surveillance. Overall, however, followup times did not differ across arms with mean followup of approximately 15 years in both groups.¹¹² The rate ratio moved toward null (RR 1.06 [95% CI, 0.87 to 1.30]) from 1.18 during the trial period. There was also no difference in ovarian cancer specific survival by arm in the trial or its extended followup (p=0.16).

An additional supplemental analysis of PLCO data aimed to determine whether use of the ROCA algorithm on CA-125 measurements collected for the PLCO would have had better performance for identifying ovarian cancer cases compared with the PLCO screening protocol.¹¹³ The analysis employed a best-case scenario assumption biased toward finding a ROCA test benefit over the PLCO protocol. Namely, all cancers in the trial that would have had a positive ROCA screening test occurring earlier than a positive screening test with the PLCO protocol were assumed to have avoided mortality from ovarian cancer. Modeling with PLCO data suggested that application of the ROCA algorithm used in the UKCTOCS would have led to earlier diagnosis of cancer in 32 percent of cases that were detected using CA125 screening using a single cutoff.¹¹³ This analysis obtained a mortality relative risk of 0.90 (95% CI, 0.69 to 1.17), and the authors concluded that even under the most lenient assumptions, the ROCA algorithm would not necessarily have resulted in a beneficial trial finding.

Another recent secondary analysis of PLCO participants with a family history of breast or ovarian cancer found a non-significant trend towards diagnosis of stage I or II cancers in the screened arm compared with the usual care arm (29% versus 17%; p=0.085)¹¹⁴ and improved survival in these patients with ovarian cancer detected by screening compared to usual care; however, this apparent improvement in survival did not result in improved ovarian cancer mortality.¹¹⁴

The control condition was described as “usual care” in the U.S. PLCO and as “no screening” in the U.K. trials. If usual care included any practices that might affect ovarian cancer detection and treatment, this could potentially reduce differences between the study arms in the PLCO.

Surveys have found that a majority of U.S. primary care and reproductive specialty clinicians conduct bimanual pelvic examinations as part of their routine gynecological care for women, believing it to be an effective way to screen for ovarian cancer.^{69, 115} During the early years of the PLCO, the screening protocol included ovarian palpation with bimanual pelvic examinations, but this element was dropped from the intervention protocol 5 years into the trial. Estimates of the test performance of the bimanual examination derived from the PLCO and systematic reviews of the effectiveness of routine pelvic examinations¹¹⁶ have found limited evidence on its effectiveness, and have brought to light its poor accuracy. Thus, it is unlikely that ovarian palpation in the screening or control arm of the PLCO would influence the results observed. Essentially, women in both arms of the PLCO likely received pelvic examinations with ovarian palpation over the course of the study, and the CA-125 and TVU screening intervention can be validly compared against usual care without routine CA-125 and TVU screening. Overall rates of TVU and CA-125 testing contamination in the control arm were similar for the PLCO and U.K.-based studies, however, suggesting that differences in the control condition did not contribute to widely divergent practices in the control arms.

Implications of Stage Shift and Treatment Findings in the Absence of a Mortality Benefit

As a contextual question in this review, we examined included studies for evidence of a cancer stage or type shift We focused the evaluation of stage shift on comparisons in the trial arms between women diagnosed with localized disease (stage I) and those with regional or distant disease (stages II-IV). These comparisons are relevant because of the higher survival rates associated with disease diagnosed at the localized stage and clinical differences in treatment strategies between stage I and higher-stage ovarian cancer (i.e., need for adjuvant radiation therapy).¹ Patients with stage II ovarian cancer have high recurrence rates. As a result, in 2009 the Gynecologic Oncology Group has recommended that stage II patients be included in trials that assess treatments for advanced-stage disease.¹ Tests for differences in proportions were conducted in Stata.

Detection of a higher proportion of localized cancers in the screening arms compared with control arms was reported in the UKCTOCS.⁹² A statistically significant (p<0.005) greater proportion of cases was identified at the localized stage (stage I) in the CA-125 ROCA (36%) and ultrasound (31%) arms than in the control arm (23%) (calculated from reported data on cancer stage). The overall differences by arm and stage were also statistically significant when comparing localized and regional cancers (stages I and II) to more advanced stages (stages III and IV). Differences between the CA-125 ROCA and ultrasound arms were not statistically significant.⁹² In the PLCO⁹¹ there was not a statistically significant difference in the proportion of cases identified at the localized stage in the intervention versus usual care group (15% versus 10%, p =0.08) and comparisons were not statistically different when comparing localized and regional cancer cases to more advanced cancers. In addition, an analysis of data from PLCO found that, compared with the usual care arm, fewer cancers in the intervention arm were categorized as type II ovarian cancers. Additionally, within the screening arm, fewer type II cancers were screen-detected. Overall, type II tumors were less likely diagnosed from screening and were diagnosed at a later stage. The authors suggested that overdiagnosis of more indolent cancer types could, in part, account for the lack of a mortality benefit from ovarian cancer screening in the trial.¹¹⁷

Only the UKCTOCS reported cancer cases by type and study arm, but did not conduct statistical tests of type differences. A greater proportion of cancers occurring in the control condition were type II cancers, which tend to be more aggressive; 60 percent cases in the TVU and CA-125 ROCA screening arms were type II compared to 64 percent in the control arm.⁹² Conversely, more cases in the screening arms were non-epithelial and borderline cancer types, which tend to progress more slowly or remain indolent; 16 percent in the CA-125 ROCA arm, 20 percent in the TVU arm, and 11 percent in the no screening arm. Borderline and non-epithelial types are more likely to be early stage cancers, and have high rates of survival.⁷ Thus, detection of more cases of indolent disease with screening may have had limited impact on ovarian cancer survival, even though it appears to contribute to an observed shift in disease stage. Among invasive epithelial and peritoneal cancer cases found, however, the CA-125 ROCA screening test appeared to identify slightly more cases at an earlier stage. Notably, however, even stage I cancers in some type II high grade epithelial carcinomas may be associated with microscopic metastases, as cancer cells can be present in ascites (stage Ic).^{2, 3, 5, 6} In addition, there were no differences in treatment approach by study arm in the PLCO; 81 percent in the intervention group received surgery plus systemic therapy, compared with 80 percent in the usual care group.⁹¹ Treatment outcomes for participants in the UKCTOCS⁹² have not yet been published. While there is some evidence of a stage shift for the CA-125 ROCA and ultrasound intervention arms in the UKCTOCS, this shift did not confer a statistically significant mortality benefit.

The absence of a mortality benefit in large, well-conducted trials has contributed to concerns the most common cancer type spreads so rapidly that it cannot be identified at an early enough stage to substantially shift mortality. The stage shift in UKCTOCS trial would seem to counter this, but the lack of mortality benefit may indicate that the early stage tumors detected are more aggressive tumor phenotypes with poor prognosis even when identified at earlier stages. Recent work to refine the distinctions among ovarian cancer molecular, pathological, and clinical characteristics highlight this point in noting that survival differences are more likely attributable to type than to stage at diagnosis, with the most common type II cancers being particularly lethal regardless of stage, likely owing to microscopic lesions that are not detectable before significant spread has occurred.⁶

Discussion of Harms

The UKCTOCS employed a more nuanced approach to CA-125 testing and triage by using an algorithm that incorporates CA-125-level trajectories, assigning three levels of risk to direct surveillance and triage tests. This was aimed at reducing rates of surgical investigation, and indeed surgery rates were lower in the CA-125 ROCA arm than in the ultrasound-only arm of the trial. Accordingly, false positive surgery rates in the CA-125 ROCA arm of the UKCTOCS were markedly lower than in the PLCO (1% versus 3%).

The surgical complication rates differed considerably for the PLCO and UKCTOCS, with 15 percent of women who underwent false positive surgery experiencing a major complication in the PLCO and just over 3 percent having a major complication from false positive surgery in the UKCTOCS. Differences in the study settings could account in part for this difference, as diagnostic testing in the PLCO was conducted through referrals to women's routine sources of care, and not necessarily specialized tertiary care settings. In contrast, in the U.K. all women referred for diagnostic testing were seen at National Health Services tertiary care surgical centers. It is unclear whether the complication rates observed in the PLCO would be observed in current U.S. community surgical practices for women referred for diagnostic testing from primary care. Current U.S. data on the complication rates for diagnostic oophorectomy (alone) are not available, but it is likely that rates vary by setting, region, and clinician characteristics.¹¹⁸ Regardless of the complication rates, high rates of surgery and removal of an ovary or ovaries in the absence of disease occurred in both trials, although rates of surgery were lowest in the CA-125 ROCA screening intervention in the UKCTOCS trial.

False positive surgical investigations in the included trials were reported to often include bilateral salpingo-oophorectomy (BSO), based on common practice for the investigation of suspected ovarian cancer.³² Given that these screening interventions were undertaken in healthy women, potential harms of unnecessary removal of the ovaries in postmenopausal women deserves scrutiny. A recent systematic review evaluated the general health consequences of BSO at the time of hysterectomy for benign indications.¹¹⁹ Health consequences for women undergoing surgical investigation for a false positive ovarian cancer screen receiving BSO would likely be comparable. Although evidence was somewhat limited, the review found reductions in ovarian cancer and in rates of reoperation for women who underwent BSO, particularly for women younger than 45 years old. There was also evidence of potential adverse effects on cardiovascular health and all-cause mortality. Sexual function may also be negatively impacted by BSO conducted at the time of benign hysterectomy.¹¹⁹ Thus, the removal of ovaries and or fallopian tubes at the time of a surgical investigation for a false positive screening test result may have downstream harms beyond those owing to the direct effects of surgery on health outcomes. Evidence on the effects of BSO among older postmenopausal women similar to those included in this trial is limited, and firm conclusions about adverse or beneficial health or social effects are not possible.

Limitations of the Review

Although a body of evidence on the test performance of various screening strategies exists, the most promising approaches using ultrasound and CA-125 have been assessed in trials. Our review did not consider observational evidence, where some tools have appeared promising in early investigations.

Given the rarity of ovarian cancer, and the invasiveness of diagnostic surgery for positive screening results, the mortality reduction from screening relative to an unscreened group is key evidence for this condition, as it summarizes the net effect of screening, detection, and treatment.¹²⁰ An effective ovarian cancer screening program among asymptomatic average-risk women would be hypothesized to save lives through lower rates of death from ovarian cancer. Observation of a cancer stage shift toward more localized cancers may lead to less morbid treatments and could underlie an observed screening mortality benefit, but stage shifts do not necessarily confer a mortality benefit. Namely, a screening test might identify cancers that would not have progressed or earlier treatment might not sufficiently change the survival rates of women to make a difference relative to an unscreened group. Even in the absence of an organized screening program, asymptomatic ovarian cancers may be detected opportunistically. Evidence from randomized trials of ovarian cancer-screening programs that report mortality outcomes can establish whether specific screening protocols result in better health outcomes (e.g., reduced mortality) than usual care or the absence of the screening program. This direct evidence is available from two large trials, and neither provides evidence of a screening benefit with any of the screening protocols tested.

Given the low incidence of ovarian cancer, very large trials are necessary to determine whether the benefits of a screening program outweigh the harms of diagnostic testing, which for ovarian cancer necessarily involves surgery and ovarian removal. We are confident that our review identified all relevant trials with ovarian cancer-mortality outcomes reported. Two additional trials that did not meet our inclusion criteria on the basis of study design and outcome reporting, but enrolled large samples of women to ovarian cancer screening would not have changed our findings had they been included.

The Shizuoka Cohort Study of Ovarian Cancer Screening and Shizuoka Cancer Registry (SCSOCS trial) randomized asymptomatic postmenopausal women in Shizuoka, Japan, to screening using ultrasonography and CA-125 (with a cutoff of 35 U/ml) (n= 41, 688) or followup without screening (n=40,799).¹²¹ Screenings were repeated yearly for an average of 5.4 screens and a mean followup of 9.2 years. Information on the impact of screening on ovarian cancer diagnosis was published in 2008, with no significant difference between the number of women with ovarian cancer detected between the screening and control groups. Differences in the percent of patients detected with stage I cancer were also not statistically significant (63% versus 38%, p=0.23). Mortality data have not been published from this trial, and it is unclear if additional analyses are planned.¹²¹ The SCOSCS was not included in the results of this review because health outcomes (e.g., mortality) have not been reported for this trial.

The University of Kentucky Ovarian Cancer Screening Trial was initiated in 1987 to assess the use of annual transvaginal ultrasound to detect ovarian cancer in asymptomatic women aged 50 or older and women aged 25 or older with a family history of ovarian cancer. ¹²² The study was a controlled trial, but did not randomize women to a control condition, instead comparing the screened cohort to a cohort of women with ovarian cancer outcomes reported in the Kentucky Tumor Registry. Data published on the effect of screening TVU on over 37,000 women from 1987 to 2011 reported that in the screening cohort 47 percent of cancers were detected at stage I and 70 percent were detected with stage I or II disease. In contrast, only 27 percent of those entered into the Kentucky Tumor Registry during the same time period had stage I or II disease (p<0.01). The 5-year disease-free survival rate for those cancers detected by screening was 85 percent compared with 54 percent of unscreened women treated at the same cancer center using the same surgical and chemotherapeutic protocols (p<0.001).¹²² This study also reported fewer surgical complications (10% of subjects) compared with the PLCO.¹²³ Our review did not include this evidence because the screening and comparison cohorts included both average-risk and high-risk women, and the participants were not randomly assigned to study groups. Further analyses of this investigation have not been reported, and there are limits with regard to conclusions that can be drawn about the effectiveness of a TVU screening program based on a cohort comparison study design.

The scope of this review was limited to the type of evidence that would be necessary to inform a change in clinical practice in accordance with USPSTF standards. While some topics evaluate the effectiveness of screening through an indirect pathway logic model, considering the performance of a screening test separately from the effectiveness of treatment. Evidence that tests the effect of screening compared with the absence of screening on intended health outcomes in a randomized design does not require as much inference across heterogeneous bodies of evidence. Nevertheless, trial evidence can have limitations in terms of generalizability and applicability to usual care. The PLCO⁹¹ is more applicable to a U.S. setting than the UKCTOCS,⁹² since the PLCO referred women to usual care settings. The low surgical complication rates from surgery seen in the UKCTOCS⁹², for example, may have been due to the receipt of care in tertiary care centers which is standard in the U.K. health system. Similarly, screening tests offered through a trial might be more accurate than screening performance in routine care settings, or surgical investigations might be more common in the absence of trial protocols.

Future Research Needs

Given null findings from two major, well-powered trials, future research may focus on identifying women at elevated risk of ovarian cancer that could benefit from screening or prophylactic interventions. More work is needed, however, to develop approaches for assessing family history and ovarian cancer risk in primary care and to optimize women's use of genetic counselors. Research is also needed to identify new markers with greater sensitivity and specificity for detection of ovarian cancer in average-risk women. In 2016, the IOM recommended additional research should focus on the development and assessment of early detection strategies that extend beyond current imaging modalities and biomarkers and reflect the pathobiology of each ovarian cancer subtype.²

The appearance of diverging cumulative mortality curves in the later years of the UKCTOCS has been a focus of the study investigators and critical commentaries. Given the absence of any effective screening modality to reduce mortality from ovarian cancer, any hint at a possible benefit merits close attention. As the UKCTOCS investigators point out, a few more years of data from the trial are needed to accrue followup data on women who entered the trial later in the enrollment period so that the complete findings are included in the later years of the analysis. Nevertheless, questions can be asked in the meantime about the mechanism that might underlie a screening intervention benefit for ovarian cancer occurring several years after the screening program ended. Given the natural history of ovarian cancer, it is unclear how a screening intervention aimed at identifying ovarian cancer and intervening at a more treatable stage would have a delayed effect. One possible explanation is that the screening activities resulted in the removal of the abnormal appearing but nonmalignant ovaries and fallopian tubes of women, and some of these otherwise might have gone on to develop ovarian cancer in later years. Given the relatively high false positive surgery rates seen in this trial, it is possible that prophylactic removal of selected women's ovaries and fallopian tubes would have effects on the later divergence of the cumulative mortality curves. A sensitivity analysis of the trial data using ovaries at risk, rather than women at risk as the dominator, might shed light on this potential effect on the trial results. The absence of an overall difference in ovarian cancer incidence suggests that the higher rate of oophorectomy may not be influential, but analyses examining the cumulative ovarian cancer incidence over the course of the trial and with more followup data could be informative.

More detailed data on the surgeries following screening would shed further light on possible interpretations of the diverging cumulative risk curves highlighted by the study authors. While rates of ovarian cancer have been found to be lower among women opting for removal at the time of hysterectomy for benign pathology or elective reasons,⁵³ reductions in the risk of rare ovarian cancers have to be weighed against possible negative effects of oophorectomy for other conditions, and risks associated with surgery.^{119, 124, 125}

Evidence of health risks associated with removal of the ovaries, although limited among post-menopausal women, includes negative consequences for cardiovascular health, sexual function, and some mental health outcomes.¹¹⁹ Given growing recognition that many ovarian cancers originate in the fallopian tubes, prophylactic salpingectomy with ovarian conservation at the time of surgery for other indications is viewed as a potentially promising preventive strategy for women undergoing pelvic surgery for other reasons. Overall, in the United States, rates of opportunistic salpingectomy are increasing, although opportunistic BSO is far more common.¹²⁶ Nevertheless, ACOG does not endorse routine salpingectomy at the time of benign indication hysterectomy or sterilization until rigorous observational or trial evidence on this intervention is available.⁵¹ Data from a large community-based health system in Northern California reported an increase from 15 percent to 72 percent between 2011 and 2014 in the practice of salpingectomy at the time of elective hysterectomy for benign indications following the distribution of resources related to the potential benefits of the practice.¹²⁷

Differences in the oophorectomy rates were also estimated in the PLCO, based on a supplemental survey following the screening period (66% of participants responded).^{91, 101} More women in the intervention arm reported an oophorectomy (7.7% versus 5.8%, RR 1.33 [95% CI, 1.24 to 1.43]). The PLCO did not find a long-term benefit of screening; the cumulative mortality effect estimate was in the direction of harms rather than benefits of screening. In the longer-term followup data from the trial, however, the intervention effect estimate moved closer to null.¹¹²

We identified no ongoing randomized trials of ovarian cancer screening using new screening tools. While some tools in development may hold promise for the future (e.g., microRNA),² currently there are no new screening tools (i.e., biomarkers, instruments) exhibiting levels of test performance beyond what is observed for the screening tools evaluated in trials.

The UKCTOCS trialists are engaged in efforts to improve upon the ROCA algorithm, adding other protein markers along with CA-125 to new prediction models derived using data from the UKCTOCS data. These models would require further validation and testing to ascertain whether they truly represent improvements on the ROCA algorithm that would potentially attain clinical benefits for ovarian cancer detection and treatment. In any case, given the absence of a single marker or screening device that is effective for ovarian cancer, research is likely to increasingly aim to identify new markers and combinations of markers in prediction models.¹²⁸

Methods for identifying women at high-risk for ovarian cancer may help to direct preventive interventions. In 2016, a report from the Institute of Medicine recommended that research focus on developing and validating dynamic risk assessment tools for identifying those at high risk of cancer, including increasing rates of genetic counseling, and to exam the risk-benefit balance of nonsurgical and surgical prevention strategies in these populations.²

Conclusion

Since the previous review, results from a large trial conducted in the United Kingdom were published. Ovarian cancer mortality did not differ between control and intervention screening conditions in any of the included trials, including two good-quality studies with adequate power to detect differences. Harms of screening include surgery resulting from a false positive. These surgeries often result in the removal of one or both ovaries and/or fallopian tubes, and can lead to major surgical complications. Reports from the UKCTOCS of a potential delayed effect of screening on ovarian cancer mortality require further follow data to evaluate, but the causal mechanism for a delayed screening effect is unclear. Major trials of promising ovarian cancer screening tools have null findings to date among healthy average-risk women, and there are considerable harms associated with screening. Further analysis of existing trials and research on new biomarkers, new risk-assessment tools, and trials of prophylactic interventions may ultimately be found to be useful in reducing ovarian cancer mortality and will need to be weighed against known screening harms.