Comparison of Withdrawal Rates in the Screening Groups
Of women with repeat testing, 12.9% (1560/12,073) withdrew in the MM arm and 20.1% (1939/9660) withdrew in the US arm. This led to an estimated relative risk of withdrawal of 1.44 (95% CI, 1.34–1.54; P ≤ 0.001) for the US arm vs the MM arm. Adjustment for center, through stratification, and age, through the inclusion of a continuous age variable, standardized at age 61 years, in the model, led to a relative risk estimate of 1.46 (95% CI, 1.36–1.56; P < 0.001) for the comparison of US and MM arms in this sample.
Effects of other Factors
Tables 3 and 4 show the times to withdrawal of participants and withdrawal rates by year of entry into the events sample. Table 5 presents the results of the analyses. In single-factor analyses after adjustment for center, age, and randomization group, GHQ ≥ 4, current STAI state score (standardized at mean 35.9), a high STAI trait value, the level of the women’s last screen, and the maximum screening level experienced, along with randomization to the US arm, demonstrate very significant (P < 0.001) relationships to withdrawal rate. Other weaker but significant relationships were observed including having a perceived chance of OC of 1 in 10 (P = 0.01), having used hormone replacement therapy (HRT; P = 0.03), and with the year of the women’s first screening event (P = 0.01). There was no evidence of an ethnicity (white vs nonwhite) effect.
Multivariate analyses demonstrate that GHQ ≥ 4 and current STAI score retain their relationship when STAI trait is included in the model. High STAI trait retains a significant but less strong relationship to withdrawal when GHQ ≥ 4 is in the model, but low STAI trait does not. However, low STAI trait is significant and high STAI trait is not significant if STAI state is in the model and not GHQ ≥ 4. When both GHQ ≥ 4 and STAI state are in the model, then the significance of STAI trait dominates that of GHQ ≥ 4. There is also a suggestion that some increase in withdrawal rate may be associated with a low STAI trait (P = 0.10).
Also maintained in the multivariate analysis is the finding that an increased risk of withdrawal is associated with a woman last having level 1 or level 2 screens. However, if a woman has previously had a level 2 screen, it brings a reduced risk of withdrawal. It seems therefore that a recent event does increase the chance of withdrawal, but if a woman had a level 2 screen previously and stayed in the program, then subsequently the risk of withdrawal is reduced. There is no evidence for an interaction between past screening level and current level. Thus, relative to a woman whose last screen was an annual screen and who only ever had a level 1 screen previously, the risk of withdrawal is estimated to be increased by a factor of 1.32. For women who last had a level 1 and level 2 screen, respectively, it increased by a factor of 1.49. However, if a the previous screen was at level 2, the rate of withdrawal is estimated to be decreased by factors of 0.56, 0.74, and 0.83 if the woman last had an annual, a level 1, and a level 2 screen, respectively. Also maintained in the multivariate analysis is an increased risk of withdrawal for women who took HRT and a decreased risk the longer the woman has been in the screening program prior to their first screening event.
If interactions between randomization group and past screening level and between randomization group and current screening level are added to the multivariate model, then the former is highly significant (P < 0.001) and the latter has a suggestive significance level of P = 0.055. The estimated relative risk associated with level 1 and level 2 screen are 1.15 (CI, 0.94–1.40) and 1.35 (CI, 1.05–1.73), respectively, for the MM group but have the higher values of 1.54 (CI, 1.29–1.84) and 1.54 (CI, 1.31–1.80) for the US group. The estimated relative risk for a previous level 2 screen is 0.82 (CI, 0.67–1.00) and 0.46 (CI, 0.40–0.55) in the MM and US groups, respectively. There was no evidence of an interaction between randomization group and ethnicity (P = 0.37).
If the interaction between STAI state and last screen level is introduced into the multivariate model (c), then there is marginal evidence for an interaction (P = 0.07 on 2 df test). If a model is fit to the groups’ last level 1 and last level 2 classes together, then the interaction between this variable and STAI state is more significant (P = 0.026 on 1 df test). The magnitude of the interaction effect is relatively small with the estimated effect of a level 1 or level 2 screen corresponding to a relative risk of withdrawal of 1.38 for a STAI state score at its mean of 35.9 but of 1.51 if the STAI state score is 10 units higher.
Although not included in the multivariate model because it is potentially on the causal pathway to withdrawal, women also answered a question about the likelihood of returning for the following year’s screen. Available responses were “yes,” “unsure,” and “no.” As expected, the withdrawal rates in the next year varied greatly with the response to this question, with rates of 3% (2600/83,508), 23% (332/1447), and 41% (186/453) observed over all patient-year observations. A formal analysis of time to withdrawal, adjusting for group, age, and center leads to a relative risk of 5.4 (CI, 4.8–6.1) and 4.9 (CI, 4.1–5.7) for the “unsure” and “no” groups relative to the “yes” group. The unadjusted relative risk estimates were 6.1 (CI, 5.5–6.9) and 5.2 (CI, 4.5–6.1).
Establishing the sensitivity and specificity of different types of screening is vital before initiating a population screening program, but it is also crucial to ascertain all potential factors that might influence participation and regular attendance. For example, a survey examining the acceptability of TVS scan within UKCTOCS showed that most women (72%) did not rate the procedure painful and only a small proportion (4%) felt embarrassed during the scan, but most shared apprehensions about the intrusive nature of the test. However, those who did rate TVS as uncomfortable and painful were less compliant with attendance at the following year’s scan compared with those who did not experience pain.2
The findings from this UKCTOCS analysis suggest that postmenopausal women who have experienced a screening event are more likely to withdraw from OCS programs using more invasive screening procedures such as TVS. High trait anxiety and increased psychological morbidity significantly influenced the rate of withdrawal, even when age, screening center, and screening group were taken into account. It is recognized that people who are anxious about their health are more likely to misinterpret health information; the likelihood of withdrawal from UKCTOCS increased in women with the incorrect perception that their chance of developing OC was as high as 1 in 10. The cognitive-behavioral theory of health anxiety predicts atypical responses in health anxious individuals when exposed to health-related information.15 Information is more likely to be viewed as personally threatening, and they are less likely to be reassured by medical investigations. This has been shown in a study of colorectal cancer screening. Health anxious people reported lower levels of reassurance after a clear test result than nonanxious participants, although the size of the effect was small.4
In the context of UKCTOCS, there was some evidence that a volunteer who was more anxious before annual screening and who then required additional screening because of a false-positive result was also more likely to withdraw from the program compared with someone less anxious having a false-positive result. As the experience of a “near miss” might heighten awareness of cancer and illness and increase anxiety, one might predict that an already anxious woman experiencing a number of false-positive results would be more likely to withdraw from screening. This supposition was not upheld by our results.
In an earlier study that examined the psychological sequelae associated with abnormal screen results for women in UKCTOCS, screening did not seem to raise anxiety, but psychological morbidity was slightly elevated by more invasive testing after annual screens. Anxiety in fact decreased with every year a volunteer stayed in the study.3 Likewise, in this withdrawal analysis, women were less likely to leave the study the later they experienced a repeat screening event during the 7 years of screening.
In the Psychological aspects of Familial Ovarian Cancer Screening Study, the main reasons for withdrawal from OCS before surgery were previous experience of UK FOCCSS Phase 1 screening, repeat tests during previous screens, higher cancer-specific distress, and a belief that aging caused OC.16 In this situation, where there is a familial risk of developing OC, it could be that having repeat tests heightened anxiety to such an extent that women decided to undergo risk-reducing surgery.
Similarly in the case of breast cancer screening programs, a false-positive result brings with it several disadvantages to the participant and provider. There is the cost related to the provision of further tests, biopsies to deliver a diagnosis, and the anxiety experienced by the participant that would never have happened in the absence of screening.4 An earlier study reported that despite having received a final clear result during routine breast screening, women who had undergone further investigations, for example, fine needle aspiration, surgical biopsy or been placed on early recall, suffered significantly greater adverse consequences at 1 month before their next routine breast screening appointment than did women who had received a clear result after their initial mammogram at their last routine breast screening.6 The authors concluded that undergoing further investigations did not necessarily motivate women to attend for their next routine appointment. Whether these women had a high anxiety trait characteristic is unknown.
Strengths and Limitations
The primary strengths of the UKCTOCS psychosocial study are its size and, unlike much previous research in this area, its longitudinal design. Cross-sectional work does not permit the prospective detailed examination over time of putative psychosocial factors that might influence withdrawal from a screening program after at least 1 repeat screen. A limitation of the study is that those women who consent to participate in a trial of different screening modalities, which also included a control group, might be a self-selected population; however, the sociodemographic characteristics of the 202,638 volunteers in UKCTOCS seem to be wide ranging.
There are no other OCS studies of comparable design and size with which to compare these UKCTOCS withdrawal results. What we have shown is that women with a high predisposition toward anxiety are more likely to drop out of screening, as do those who experience high anxiety after their most recent scan. Furthermore, the more invasive the initial screening procedure is, that is TVS, rather than a multimodal approach, the more likely withdrawal will be after a repeat scan or false-positive result.
The United Kingdom Collaborative Trial of Ovarian Cancer Screening included a comprehensive psychosocial arm that has permitted an in-depth appraisal of not only the psychosocial harms and benefits of OCS but also some of the factors that might enhance or inhibit attendance and re attendance. Next year, the National Screening Committee is scheduled to review its policy on OCS in women after the UKCTOCS study against criteria that include psychosocial factors. These results should assist policy makers when considering the optimal screening methods and any accompanying educational resources, especially aimed at ameliorating anxiety.
The authors thank the many volunteers who participated in the trial, the clinicians and research nurses in the 16 centers, data monitors, managers, and other academic staff involved over the years, especially staff at SHORE-C, University of Sussex. The authors appreciate the help of UKCTOCS staff at UCL and the helpful comments from the UKCTOCS trialists and steering committee.
2. Gentry-Maharaj A, Sharma A, Burnell M, et al. Acceptance of transvaginal sonography by postmenopausal women participating in the United Kingdom Collaborative Trial of Ovarian Cancer Screening. Ultrasound Obstet Gynecol
. 2013; 41: 73–79.
3. Barrett J, Jenkins V, Farewell V, et al. Psychological morbidity associated with ovarian cancer screening: results from more than 23 000 women in the randomised trial of ovarian cancer screening (UKCTOCS). Br J Obstet Gynaecol
. 2014; 121: 1071–1079.
4. Miles A, Wardle J. Adverse psychological outcomes in colorectal cancer screening: does health anxiety play a role? Behav Res Ther
. 2005; 44: 1117–1127.
5. Hofvind S, Thoresen S, Tretli S. The cumulative risk of a false-positive recall in the Norwegian Breast Cancer Screening Program. Cancer
. 2004; 101: 1501–1507.
6. Brett J, Austoker J. Women who are recalled for further investigation for breast screening: psychological consequences 3 years after recall and factors affecting re-attendance. J Public Health Med
. 2001; 23: 292–300.
7. Taylor KL, Shelby R, Gelmann E, et al. Quality of life and trial adherence among participants in the prostate, lung, colorectal and ovarian cancer screening trial. J Natl Cancer Inst
. 2004; 96: 1083–1094.
8. Klompenhouwer EG, Duijm LEM, Voogd AC, et al. Re-attendance at biennial screening mammography following a repeated false positive recall. Breast Cancer Res Treat
. 2014; 145: 429–437.
9. Menon U, Gentry-Maharaj A, Ryan A, et al. Recruitment to multicentre trials—lessons from UKCTOCS: descriptive study. BMJ
. 2008; 337: 1283–1286.
10. Menon U, Gentry-Maharaj A, Hallett R, et al. Sensitivity and specificity of multimodal and ultrasound screening for ovarian cancer, and stage distribution of detected cancers: results of the prevalence screen of the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS). Lancet Oncol
. 2009; 10: 327–340.
11. Fallowfield L, Fleissig A, Barrett J, et al. Awareness of ovarian cancer risk factors, beliefs and attitudes towards screening: baseline survey of 21715 women participating in the UK Collaborative Trial of Ovarian Cancer Screening. Br J Cancer
. 2010; 103: 454–461.
12. Spielberger C, Gorsuch R, Lushene R, et al. Manual for the State-Trait Anxiety Inventory (Form Y1-Y2)
. Palo Alto, CA: Consulting Psychology Press; 1983.
13. Goldberg D. Manual of the General Health Questionnaire
. NFER-Nelson: Windsor; 1978.
14. Cox D. Regression models and life tables (with discussion). J R Stat Soc Ser B
. 1978; 34: 187–220.
15. Hadjistavropoulos HD, Craig KD, Hadjistavropoulos T. Cognitive and behavioural responses to illness information: the role of health anxiety. Behav Res Ther
. 1998; 36: 149–164.
16. Lifford KJ, Frazer L, Rosenthal AN, et al. Withdrawal from familial ovarian cancer screening for surgery: findings from a psychological evaluation study (PsyFOCS). Gynecol Oncol
. 2012; 124: 158–163.
Keywords:© 2015 by the International Gynecologic Cancer Society and the European Society of Gynaecological Oncology.
Ovarian cancer; Population screening program; Withdrawal; Anxiety; Psychological morbidity; UKCTOCS