Compliance of Average- and Intermediate-Risk Women to Semiannual Ovarian Cancer Screening

Screening to detect early-stage ovarian cancer has potential for improving survival. Cure rates for early-stage disease are nearly 90%, yet overall ovarian cancer survival is only 50% because the majority of women have advanced-stage disease at the time of clinical diagnosis (1). Understanding factors that affect compliance and adherence to ovarian cancer screening protocols is critical if ovarian cancer screening is to be effective in reducing overall ovarian cancer mortality.

Ovarian cancer is relatively uncommon, affecting roughly 24,000 women in the US each year. A positive family history of breast and/or ovarian cancer is the strongest risk factor for the disease yet this accounts for only 4–6% of cases (2, 3). An individual's risk for developing cancer can span decades (4). Consequently, to detect the majority of ovarian cancer cases, most women need to be screened and strict adherence to screening protocols over time will be essential.

Screening strategies using transvaginal sonography (TVS) alone or in combination with serum CA125 are able to detect pre-clinical ovarian cancer. Annual TVS is nearly 100% sensitive but results in approximately 12 surgeries for each cancer case detected (5). Screening using a CA125 threshold of 30 mIU/l to refer women for TVS is 80% sensitive, has a positive predictive value of 21%, and, in a pilot randomized trial of 22,000 women, improved the median survival of women with ovarian cancer by over 30 months (6). Several large-scale randomized controlled trials designed to evaluate the impact of ovarian cancer screening on mortality are now underway (6, 7).

Prior studies with a focus on high-risk or self-selected women involved in annual screening have identified educational background, distance from the screening center, ovarian cancer risk, and ovarian cancer worry as factors affecting participation in or compliance to ovarian cancer screening protocols (8, 9). It is not clear whether these findings pertain to women who are not high-risk or women screened more intensively.

This paper reports on the participation of average- or intermediate-risk women randomly assigned to a semiannual ovarian cancer screening program and evaluates the impact of demographic, psychological, and health-related factors on screening compliance and adherence. We address the implications of our findings for the effectiveness of ovarian cancer screening in general.

We enrolled 592 cancer-free women, aged 30 and older, at average or intermediate risk for developing ovarian cancer into a 2 × 2 factorial, randomized controlled trial evaluating the quality-of-life effects of ovarian cancer screening and ovarian cancer risk education sessions alone or in combination (the QUEST trial). Participants were recruited from (a) mass mailings from a purchased mailing list (n = 397) which included women from selected zip codes located in the vicinity of the study center; (b) a list of volunteers who had participated in an annual fundraising event for an ovarian cancer research center in 1996, 1997, or 1998 (n = 132); (c) volunteers who contacted the center asking about ways to participate in research projects (n = 30); and (d) and physician referral (n = 33). In this paper, we report a secondary analysis of data limited to the 292 women randomized to study arms that included ovarian cancer screening. Of these women, 140 were randomized to screening alone and 152 were randomized to screening plus risk education sessions.

Participant risk for developing ovarian cancer was determined by responses to questions about personal and family history of cancer contained in a pre-enrollment questionnaire. This included a series of questions about breast, ovarian, colon, and other cancers in first- and second-degree relatives. Criteria used to assign risk classification are outlined in Table 1. Women were considered high-risk if their reported family history predicted at least a 10% probability of a germline mutation in the BRCA1 or BRCA2 genes or if they met the Amsterdam criteria for the hereditary nonpolyposis colorectal cancer syndrome (HNPCC). Women were considered intermediate-risk if they reported a personal history of breast cancer or reported a single case of ovarian or two or more cases of breast cancer in the family. Women reporting lesser affected family histories were categorized as average-risk. High-risk women were excluded from enrollment in the study to avoid randomizing them to a study arm without ovarian cancer screening.

A separate baseline questionnaire was used to confirm the reported family history, to obtain demographic data, and to gain information about overall quality of life, prior ovarian cancer screening experience, perceived ovarian cancer risk, and ovarian cancer worry. Participants reported the highest level of schooling completed and rated their general health as “excellent,” “very good,” “good,” “fair,” or “poor.” They also reported whether they had previously received a CA125 blood test to screen for ovarian cancer and if they had received an ultrasound examination of their ovaries. Responses were “yes,” “no,” “don't know,” or “refused.” Women responding “no” to both questions were classified as not having been previously screened. Women responding “yes” to one or both questions were classified as having been previously screened. Individuals who responded “yes” were questioned further about the time interval from their last screen. These women were subclassified as currently being screened if their last CA125 or ultrasound was within 1 year of study enrollment. Perceived risk of developing ovarian cancer was determined by asking participants to estimate their chances of developing ovarian cancer, from 0% (no chance) to 100% (definitely will get it) and their risk of developing ovarian cancer relative to others of the same age (risk lower than most women, average risk, risk higher than most women).

Worry about ovarian cancer risk was assessed using a modified version of the Lerman cancer worry scale (10) which was developed for the purpose of assessing worry about breast cancer and has been used with women at high risk for cancer due to family history. Participants were asked about the frequency with which they worry about getting ovarian cancer and the frequency with which thoughts about getting cancer affect their mood or ability to perform daily activities. Responses to these questions were recorded using a four-point scale from “rarely or not at all” to “almost all the time.” Women's responses to these items were used to categorize women as “not worried,” “mildly worried,” “moderately worried,” or “severely worried.” This method of categorization has been used in several studies of high-risk women (11–13). Women were considered not to be worried if they reported thinking about their risk “rarely or not at all.” Women were considered mildly worried if they reported thinking about their risk “sometimes” during the last month but did not report worries affecting their moods or daily activities. Women who reported thinking about their risk “often” or “almost all the time” or who reported that their thoughts about their risk affected their mood “sometimes” were considered moderately worried. Those who reported that worry about risk affected their mood “often” or “almost all the time” or who reported that their thoughts about risk affected their daily activities “sometimes,” “often,” or “almost all the time” were considered severely worried about their risk.

The ovarian cancer screening protocol consisted of CA125 blood tests and TVS alternating semiannually over an 18-month period for each woman, spanning a period of 3 years for the entire cohort. All women were first screened with a CA125 blood test immediately following enrollment into the study. Thus, participants were scheduled to undergo CA125 at month 1 and month 12, and TVS at month 6 and month 18, following enrollment. Screening test results were characterized as normal or abnormal using standardized criteria. Participants with abnormal screening tests results were further evaluated by ultrasonography. Individuals with elevated CA125 blood tests were referred for immediate TVS and those with normal ultrasound findings returned to regular scheduled screening. All individuals with abnormal ultrasound findings underwent repeat sonography in 6–8 weeks and those with persistently abnormal ultrasounds were classified as having a positive screen. Study protocol required that these participants be removed from the screening protocol and referred to their primary care provider with a recommendation for clinical management. These women were allowed to return to screening after completing recommended work-up.

Participants could undergo blood draws for CA125 testing at any of nine locations, seven of which were inside and two of which were outside the greater Seattle metropolitan area. A standard protocol was followed at all locations for blood processing and measurement of CA125. All ultrasound examinations were performed at a single facility in central Seattle. For premenopausal women, ultrasounds were scheduled during the early follicular phase of the menstrual cycle whenever possible. All screening tests were performed at no cost to the participants. Insurance companies were billed for the costs of screening intermediate-risk participants. The Marsha Rivkin Center for Ovarian Cancer Research, a nonprofit organization dedicated to improving outcomes for women with, or at risk for, ovarian cancer, provided funds to cover costs associated with screening all average-risk participants and intermediate-risk participants without medical insurance and those whose insurance companies denied payment.

A computerized tracking system generated screening test reminder letters and monitored adherence to the screening. Reminder letters were sent 1 month before the screen due date. For TVS, women were contacted by phone 1 week after the reminder letter was sent to offer assistance in scheduling the ultrasound appointment. This was done to minimize barriers to TVS participation associated with scheduling the screen. The post-reminder-letter assistance call was not used for CA125, because of the relative logistic ease of undergoing the blood draw; women could simply visit a blood draw center anytime during business hours. For all screens, women were considered overdue 1 month after the screen due date. Follow-up calls began at that time, and consisted of four messages left at 1-week intervals, with one to two interim attempts to reach participants between messages. Follow-up messages concluded either on telephone contact with the participant or the completion of the series of reminder messages.

The ovarian cancer risk education sessions consisted of four weekly 2-h classes led by a professional facilitator. The average class size was eight women. The purpose of the classes was to provide accurate information about ovarian cancer risk and ovarian cancer screening in a supportive group setting, and to provide training on social support and stress management skills. The first class focused on ovarian cancer risk factors, reviewing and distinguishing between known and possible risk factors, providing estimates of lifetime risk in the general population, and providing a personal estimate of risk for each class participant. The second class focused on screening for ovarian cancer, including reviewing and distinguishing between screening and diagnostic tests, understanding what screening tests are available, and defining screen test parameters such as sensitivity, specificity, and positive predictive value. Participants were fully informed about the risks and benefits of screening for ovarian cancer and about the rationale behind not recommending screening for ovarian cancer outside of a research setting. The third class focused on the value of social support in coping with stress, evaluating one's own support network, and developing skills to enhance that network. The fourth class concluded the series by reviewing stress and the stress cycle, and developing skills to reduce and prevent stress. A summary of the class series was also provided at the last class.

Participants were classified as compliant to a screening test if they successfully completed the test at any time before the next scheduled screen. Participants were classified as compliant with the final scheduled screen if they completed the screen within 6 months of the scheduled date. The compliance rate for each screen was calculated as a ratio of the number of compliant participants relative to the number expected to attend a screening test. Study protocol required that participants with positive screens be temporarily removed from the screening program while undergoing clinically appropriate follow-up or management. These women were not included in the compliance rate calculation during the time in which they were removed from the screening program. Participants who successfully completed all four screening tests were classified as adherent to the screening program. Attrition from the screening program was defined as having attended all prior screens but failing to attend a scheduled screen and all subsequent screens. McNemar's test of correlated proportion was used to compare differences in rates of compliance between screens. χ² tests were used to identify factors associated with compliance to all four screens and to CA125 or TVS screens alone. Odds ratios and confidence intervals for these associations were calculated by logistic regression.

Fifty-four percent of participants were under 50 years of age. Over 90% of women were Caucasians, received at least some college education, and resided within King County. Seventy-eight percent (n = 229) of women reported their health as either very good or excellent while only 3.4% (n = 10) reported their health as either fair or poor. Thirty-six percent (n = 106) of participants were classified as having been screened for ovarian cancer before entry into the study, including 23 women who reported having received a CA125 blood test, and 94 women who reported receiving TVS. Eleven women reported being screened with both CA125 and TVS.

Twenty-two percent (n = 63) of participants were classified as being at intermediate risk for developing ovarian cancer. A family history of ovarian cancer in a single first- or second-degree relative was the most common reason for classification of participants in the intermediate-risk (n = 42) category. Sixteen women were classified as intermediate-risk based on a personal or family history of breast cancer, and 5 women had more than one risk factor. A personal history of cancer was reported by 29 women: 11 with breast cancer, 7 with cervical cancer, 7 with melanoma, and 4 with cancers at other sites.

Nineteen percent of respondents (n = 54) estimated their risk of developing ovarian cancer as higher than average. Perceived risk of developing ovarian cancer was positively associated with risk categorization based on medical risk factors (χ² value = 93.9 P < 0.0001). Sixty percent of intermediate-risk women versus 7% of average-risk women perceived their risk of developing ovarian cancer as higher than average; however, participants appeared to overestimate their risk. Thirty-one percent of participants (n = 90) estimated that they had at least a 50% chance of developing the disease including 43% (n = 27) of intermediate-risk women and 28% (n = 63) of average-risk women. Ovarian cancer worry among participants was infrequent. Only 5% (n = 15) of respondents were classified as moderately or severely worried about developing ovarian cancer.

The compliance rate at each screen decreased over time in a non-linear pattern (Fig. 1). The reduction in compliance between screens 1 and 2 and screens 3 and 4 were similar (15.5% and 18.7%, respectively) and substantially larger than either the decrease between enrollment and screen 1 (2.7%) or the decrease between screen 2 and screen 3 (4.2%). When compared to the preceding screen, the drop-off in compliance at screen 2 and 4 was statistically significant (P < 0.0001 for both analyses). One hundred seventy-two women (59%) completed all four screens and were classified as adherent to the screening protocol.

One hundred fifty-two of the 292 women in the screening program were assigned to receive risk education. Eighty-six percent of these participants attended all scheduled education classes. Participation in the risk education sessions, demographic variables, ovarian cancer risk, perceived cancer risk, ovarian cancer worry, quality-of-life estimates, pre-enrollment ovarian cancer screening, and receiving an abnormal screen test result were not associated with adherence to the screening protocol (Table 2). Similar results occurred when analyses were limited to compliance with screens involving only CA125 or TVS (data not shown).

Most noncompliance to a screening test was due to attrition. One hundred twenty women missed at least one screen, including 42 who missed only screen 4. Of the 78 remaining women, 46 (59%) dropped out of the screening program and did not return for further screening after missing a screen. The percentage of noncompliant women at each screen who did not participate in any further screening was 75% (6/8) at screen 1, 74.5% (38/51) at screen 2, and 85% (52/61) at screen 3.

Screening compliance patterns suggest that participants favored screening with CA125 blood tests. Thirty-nine of the 230 women (17%) who attended both CA125 screens did not attend any screens involving sonography, whereas only 2 of the 180 women (1.1%) who attended both ultrasound screens avoided screens involving CA125. Furthermore, 13 of 51 women (25%) who missed a TVS at screen 2 returned for CA125 testing at screen 3. This compares to 9 of 61 (15%) of women who returned for TVS testing after having missed a CA125.

Pre-enrollment screening was not associated with compliance to screening nor did it predict a preference for a specific screening modality. Women who participated in only study screens involving CA125 were neither more nor less likely than others to be classified as having been screened before enrollment with either CA125 or ultrasonography. Similarly, women who participated in only those study screens involving TVS were neither more nor less likely than others to be classified as having been screened before enrollment with either CA125 or ultrasonography.

Compliance with the semiannual screening protocol used in this study diminished rapidly. Only 64.5% of women attended the final screen scheduled 18 months after enrollment. The drop-off in compliance may be attributable to the relatively short screening interval employed. Higher compliance rates have been reported in trials using annual screening intervals. The percentage of women who attended screens scheduled 24 months after enrollment was 78.6% for annual screening trials using CA125 and 88.9% for annual screening trials using TVS (14, 15). In our study, most women who missed a screen did not return for further screening, suggesting that they may have tired of screening.

The selection of ovarian cancer screening intervals used in clinical trials has been largely empiric or adjusted to coincide with screening intervals for other cancers such as breast cancer. Most studies have employed annual screening intervals for average-risk women and more-frequent screening for women at higher risk. We selected a semiannual screening interval for use in this study because the primary aim of the QUEST trial was to evaluate the quality-of-life effects of screening and we wanted to ensure that the screening experience of women in the screening study arms was more intensive than that of control women, even if their physicians referred them for annual screening.

The optimal ovarian screening interval is unknown; however, it is plausible that screening more frequently than annually may be appropriate. In a randomized trial involving 22,000 women, Jacobs et al. (14) were unable to demonstrate a screening-related stage shift with an annual multimodal screening program. Using a micro-simulation model, we estimated a substantially greater mortality reduction (36.4% versus 26.7%) when a multimodal screening strategy is performed semiannually rather than annually (16). In a retrospective study, Horiuchi et al. (16) reported that 16 of 35 (46%) patients had normal findings on TVS performed within 12 months of a clinical diagnosis of ovarian cancer.

Analysis of compliance by screening modality suggests that women favored screening with CA125. Compared to the preceding screen, there was a statistically significant reduction in compliance for screens involving TVS (screens 2 and 4) but not for the CA125 screen (screen 3). Additionally, more women attended only screens involving CA125s than attended only screens involving TVS. Furthermore, the majority of women who missed a screen and subsequently returned to screening skipped an ultrasound but returned for CA125 testing. None of the variables we studied, including pre-enrollment screening experience with either CA125 or ultrasound, were associated with compliance to CA125 screens. It is possible that women favored CA125 screening because the test was more widely available and easier to schedule than transvaginal ultrasound or that the ultrasound procedure was viewed as more complex and invasive compared to the blood draw. Additional studies designed to identify factors affecting preferences for specific screening modalities could lead to the development of strategies for enhancing screening compliance.

Participants in screening trials are not representative of the general population and may be more likely than others to participate in screening. Pavlik et al. (17) surveyed 9,415 of 13,963 volunteer participants in the University of Kentucky Ovarian Screening Project. Compared to county-based averages, participants tended to be better educated and have greater access to healthcare. Women in the present trial were highly educated, predominately Caucasian, and in good health. Over one-third of women reported receiving ovarian cancer screening before study enrollment. It seems likely that compliance rates lower than those reported here might be expected if a semiannual screening protocol similar to the one used in this study was employed in the general population.

None of the variables we studied were significantly associated with compliance to all four of the scheduled screens or to screens involving only CA125 or only TVS. Although not statistically significant, our data suggest an association between educational background and adherence to the screening protocol and in particular suggests a negative correlation between adherence and education level of high school graduate or lower. This finding is consistent with the University of Kentucky screening program's reports that completing 12 or more years of education was correlated with overall compliance and was the most important predictor for attending a large number of screens.

Cross-sectional studies have demonstrated a positive association between ovarian cancer worry and use of ovarian cancer screening services. We reported an association between ovarian cancer screening and mild cancer-related worry in a population-based sample of 6,749 women (18) and moderate or severe levels of worry in a survey of over 3,000 women who served as the initial recruitment resource for the QUEST trial (11). Ovarian cancer worry is positively associated with use of CA125 and pelvic sonography for ovarian cancer screening among first-degree relatives of ovarian cancer patients (9). These findings have been interpreted as being consistent with either worried women seeking screening for reassurance or screening causing increased levels of worry. The distribution of ovarian cancer-related worry assessed at enrollment in the present study population is similar to that reported in our prior studies; however, we found no association between worry at the start of the study and adherence to screening. The lack of an association between ovarian cancer worry assessed at enrollment and subsequent screening adherence suggests that worry may occur as a consequence of screening, rather than worry motivating women to seek screening. An analysis of changes in worry over time among women in this study is ongoing.

The primary goals of the risk education sessions were to provide participants accurate information about ovarian cancer risk and ovarian cancer screening, and to provide social support and stress coping skills. Classes were well attended and favorably reviewed by participants. Although adherence to screening was not improved by participation in the risk education classes, it was encouraging to find that screening participation was not negatively impacted by the informational content or additional burden of attending the classes, given that the classes were candid regarding the lack of documented benefits associated with participation in screening. It is plausible that a similarly structured education class may enhance adherence if data demonstrating the effectiveness of ovarian cancer screening becomes available.

Our study demonstrates that compliance of average- and intermediate-risk women to a semiannual ovarian cancer screening program diminishes rapidly and that participants tend to be more compliant to screening involving CA125. These findings suggest that a semiannual screening program, particularly one including TVS, may be too intensive for use in the intermediate- and average-risk population.

We thank Robin L. Foster, PhD, for her thoughtful review and Laurie Dunlap for assistance in preparation of the manuscript.