Clinical trials of population screening
Two large, well-conducted randomised controlled trials (RCTs) of population screening for ovarian cancer reported no statistically significant mortality benefit at approximately 11-12 years follow-up using a priori analyses.
From 1993 to 2001, the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO) enrolled over 70,000 women in the United States aged 55 to 74 years, who were randomised 1:1 to either a no-screening control group or to screening. The screening tests were CA125 (single threshold) plus TVUS annually for 4 years, followed by annual CA125 (single threshold) only for up to a further 2 years. No statistically significant mortality benefit or stage shift was associated with screening after a median of 12.4 years follow up,12 but 14.8 false positive surgeries were conducted per screen-detected cancer. Extended follow-up to a median of 14.7 years also found no mortality benefit.13 Post hoc analysis found no mortality or stage shift benefits with screening for the more aggressive Type II cancers.14
From 2001-2005, the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS) enrolled over 200,000 post-menopausal women aged 50-74 years, who were randomised 2:1:1 to no screening or to one of two annual screening strategies: TVUS alone or ROCA triage. The latter involved CA125 evaluation using the Risk of Ovarian Cancer Algorithm (ROCA), with the results determining whether or not women received additional testing, which could be either repeat CA125 (Level I) or CA125 and TVUS (Level II).
After a median follow up of 11.1 years, a stage shift was observed in the ROCA triage screening group, with significantly more diagnoses made at Stages I and II compared with the control group.15 No mortality benefit, however, was associated with screening when analysed using the pre-specified Cox proportional hazard model using data up to 2014.
When analysed post hoc with the weighted log-rank test used in the PLCO trial, which accounts for the expected delay in mortality outcomes, a significant mortality benefit was noted for both screening groups compared with the no-screening group. However, the validity of the post hoc analyses of UKCTOCS has been challenged.16-19 The lead investigator of this trial has commented that ‘for the time being, in the absence of unequivocal evidence of a mortality benefit, large-scale population-based ovarian cancer screening programmes are not justified’.20
When classifying borderline tumours as false positives, the rate of false-positive surgeries per screen-detected cancer was 14.8 for TVUS (which is higher than the generally accepted rate of 9 per screen-detected ovarian cancer) and 2.9 for ROCA triage.15 ROCA triage was also more sensitive than TVUS but, at 74- 84%, remained below what might be acceptable for a screening test.15 No analyses by tumour type have yet been reported for this trial. Longer term follow-up of UKCTOCS is expected after further censorship in 2018 and 2024, with reporting in 2020.