/ Published in Blog

Does annual screening with Ultrasound and CA125 prevent ovarian cancer?

Ultrasound Ovarian Cancer

Ovarian cancer population screening and mortality after long-term follow-up in the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS): a randomised controlled trial

Summary

Background

Ovarian cancer continues to have a poor prognosis with the majority of women diagnosed with advanced disease. Therefore, we undertook the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS) to determine if population screening can reduce deaths due to the disease. We report on ovarian cancer mortality after long-term follow-up in UKCTOCS.

Methods

In this randomised controlled trial, postmenopausal women aged 50–74 years were recruited from 13 centres in National Health Service trusts in England, Wales, and Northern Ireland. Exclusion criteria were bilateral oophorectomy, previous ovarian or active non-ovarian malignancy, or increased familial ovarian cancer risk. The trial management system confirmed eligibility and randomly allocated participants in blocks of 32 using computer generated random numbers to annual multimodal screening (MMS), annual transvaginal ultrasound screening (USS), or no screening, in a 1:1:2 ratio. Follow-up was through national registries. The primary outcome was death due to ovarian or tubal cancer (WHO 2014 criteria) by June 30, 2020. Analyses were by intention to screen, comparing MMS and USS separately with no screening using the versatile test. Investigators and participants were aware of screening type, whereas the outcomes review committee were masked to randomisation group. This study is registered with ISRCTN, 22488978, and ClinicalTrials.gov, NCT00058032.

Findings

Between April 17, 2001, and Sept 29, 2005, of 1 243 282 women invited, 202 638 were recruited and randomly assigned, and 202 562 were included in the analysis: 50 625 (25·0%) in the MMS group, 50 623 (25·0%) in the USS group, and 101 314 (50·0%) in the no screening group. At a median follow-up of 16·3 years (IQR 15·1–17·3), 2055 women were diagnosed with tubal or ovarian cancer: 522 (1·0%) of 50 625 in the MMS group, 517 (1·0%) of 50 623 in the USS group, and 1016 (1·0%) of 101 314 in the no screening group. Compared with no screening, there was a 47·2% (95% CI 19·7 to 81·1) increase in stage I and 24·5% (−41·8 to –2·0) decrease in stage IV disease incidence in the MMS group. Overall the incidence of stage I or II disease was 39·2% (95% CI 16·1 to 66·9) higher in the MMS group than in the no screening group, whereas the incidence of stage III or IV disease was 10·2% (−21·3 to 2·4) lower. 1206 women died of the disease: 296 (0·6%) of 50 625 in the MMS group, 291 (0·6%) of 50 623 in the USS group, and 619 (0·6%) of 101 314 in the no screening group. No significant reduction in ovarian and tubal cancer deaths was observed in the MMS (p=0·58) or USS (p=0·36) groups compared with the no screening group.

Interpretation

The reduction in stage III or IV disease incidence in the MMS group was not sufficient to translate into lives saved, illustrating the importance of specifying cancer mortality as the primary outcome in screening trials. Given that screening did not significantly reduce ovarian and tubal cancer deaths, general population screening cannot be recommended.

Funding

National Institute for Health Research, Cancer Research UK, and The Eve Appeal.

Introduction

Ovarian cancer remains the most deadly of all gynaecological cancers. Most patients (58%) are diagnosed at an advanced stage (III or IV), which is associated with poor survival (5-year survival is 27% for stage III and 13% for stage IV ovarian cancer).1 The greater than 90% survival rates in women detected at stage I1 has spurred international efforts in early detection, spanning across four decades.2, 3, 4, 5, 6 All trials have used combinations of the biomarker CA125 and pelvic imaging using transvaginal ultrasound scans (TVS). Despite these extensive endeavours, to date there is no evidence that screening for ovarian cancer saves lives.7, 8, 9
In our multicentre randomised trial (UK Collaborative Trial of Ovarian Cancer Screening [UKCTOCS]), 202 638 women from the general population were randomly assigned to two annual screening groups—multimodal screening (MMS; longitudinal CA125 and second line TVS) and ultrasound screening (USS; TVS first and second-line test), and a no screening group. We reported previously (median follow-up of 11·1 years), that an absolute proportion of 13% more women with ovarian, tubal, and peritoneal cancer were diagnosed with stage I or II disease in the MMS group than in the no screening group. There was no change in stage in the USS group. There was no evidence of a reduction in disease-specific deaths in either screened group compared with the no screening group using the Cox version of the log-rank test. The observed reduction in deaths was delayed and the cumulative mortality curves appeared to be diverging at the time of previous reporting.9 Therefore, we aimed to continue follow-up and report here on the long-term mortality effects of ovarian cancer screening in UKCTOCS.

Research in context

Evidence before this study
We searched PubMed from Jan 1, 2015, to Dec 31, 2020, with no language restrictions for randomised controlled trials for ovarian cancer screening that reported mortality data. The following keywords were used to search the database: “ovarian cancer” AND “randomised controlled trial” AND “screening” AND “mortality”. We found two relevant publications. In the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS; n=202 638), at a median follow-up of 11·1 years, no significant reduction in deaths from ovarian cancer was seen in either of the screen groups (multimodal or ultrasound) compared with the no screening group. A reduction in deaths was seen but was delayed and only apparent after about 7 years. There was a suggestion that 15% fewer women in the multimodal screening group and 11% fewer in the ultrasound screening group died from ovarian cancer compared with the no screening group. Additionally, a significantly greater proportion (13%) of women with ovarian cancers in the multimodal group but not in the ultrasound group were found at an earlier stage (stage I and II) compared with the no screening group. As the data did not definitively answer the question of whether screening saved lives, follow-up was continued to gather more evidence. The Ovarian Cancer Screening arm of the Prostate Lung Colorectal Ovarian (PLCO) cancer trial in the USA is the only other large randomised controlled trial (n=78 216) to explore mortality benefit. Following extended follow-up (median 14·7 years), the trial confirmed previous findings of no ovarian cancer mortality reduction between the screen and control arms.
Added value of this study
Long-term follow-up (median follow-up >16 years after recruitment) in the largest ovarian cancer screening trial, to our knowledge, provides definitive new evidence that neither screening approaches used in UKCTOCS reduced deaths from ovarian cancer, compared with no screening. This result was despite a 47·2% increase in incidence of women with ovarian and tubal cancer diagnosed at stage I and 24·5% decrease in those diagnosed with stage IV disease in the multimodal group compared with the no screening group. Importantly, however, there was only a 10·2% decrease in overall incidence of stage III or IV disease.
Implications of all the available evidence
General population screening for ovarian and tubal cancer with either approach used in UKCTOCS cannot currently be recommended. We need a screening strategy that can detect ovarian and tubal cancer in asymptomatic women even earlier in its course and in a larger proportion of women than the tests used in the trial. Meanwhile, our results emphasise the importance of having ovarian and tubal cancer mortality as the primary outcome in screening trials.

 

Methods

Study design and participants

We did a randomised controlled trial (UKCTOCS) of postmenopausal women aged 50–74 years from the general population recruited through 13 centres in National Health Service (NHS) Trusts in England, Wales, and Northern Ireland with use of age-sex registers of 27 primary care trusts.10
We commissioned specialised software from the NHS to randomly select women aged 50–74 years and then flag them on primary care trusts’ registers and allow electronic transfer of their personal and general practice details. We then sent women personal invitations and logged replies on the trial management system. Women attended a recruitment clinic at the regional centre where they viewed an information video, completed a recruitment questionnaire, and provided written consent and a baseline serum sample. We scanned recruitment questionnaires at the coordinating centre into a bespoke trial management system.
Inclusion criteria were 50–74 years of age and postmenopausal status. Exclusion criteria were bilateral oophorectomy, previous ovarian or active non-ovarian malignancy, or increased familial ovarian cancer risk.
Ethical approval was provided by the UK North West Multicentre Research Ethics Committees (00/8/34) on June 23, 2000. All participants provided written informed consent. The trial design has been previously published and the protocol is available online.9, 10, 11, 12

Randomisation and masking

The trial management system confirmed eligibility and then randomly allocated women using the Visual Basic randomisation statement and the RND function to annual screening using the MMS or USS strategy, or no screening in a 1:1:2 ratio. The trial management system allocated a set of 32 random numbers to each regional centre, of which eight were allocated to MMS, eight to USS, and the remaining 16 to no screening. We randomly allocated each successive volunteer within the regional centre to one of the numbers and subsequently randomly allocated them into a group. Investigators and participants were aware of group allocation but members of the outcomes committee were masked to randomisation group.

Procedures

Annual screening in the MMS group used serum CA125 measurements, with the pattern over time interpreted using the risk of ovarian cancer (ROCA) calculation, 13
which identifies significant rises in CA125 concentration above baseline. On the basis of risk, women were triaged to normal (annual screening), intermediate (repeat CA125 ROCA test in 3 months), and elevated (repeat CA125 ROCA test and transvaginal USS as a second-line test in 6 weeks) risk. Annual screening in the USS group used TVS as the primary test, which was classified as normal (annual screening), unsatisfactory (repeat in 3 months), or abnormal (scan with a senior ultrasonographer within 6 weeks). In both groups, women with persistent abnormalities were assessed by a trial clinician and referred to the NHS where they underwent further investigation or surgery. We deemed women who had surgery or a biopsy for suspected ovarian cancer after clinical assessment as screen positive.
Women were linked using their NHS number to national cancer and death registration data and Hospital Episode Statistics records (appendix p 4). An additional questionnaire was sent in June, 2020, to a subset of participants who had either exited the national registries or for whom it was not possible from Hospital Episode Statistics data to ascertain if both ovaries had been removed.

Throughout the trial, we interrogated the available sources to identify women diagnosed with any of 19 International Classification of Diseases (ICD)-10 codes for possible ovarian or tubal cancer and retrieved copies of medical notes.

The only exception was women with malignant neoplasm without specification of site (ICD-10 C80), who also had another non-ovarian, tubal, or peritoneal cancer registration. Medical notes, with any mention of randomisation group redacted, were reviewed by the outcomes review committee consisting of gynaecological pathologists and oncologists (NS, RM, RW, RA, LC, AS, and KW). The outcomes review committee assigned cancer site (ie, whether ovarian, tubal, or other site using a previously audited prespecified algorithm),

Federation of Gynecology and Obstetrics (FIGO) 2014 stage, grade, morphology, ovarian cancer type, and cause of death. We defined ovarian and tubal cancer using the WHO 2014 classification and death due to ovarian and tubal cancer based on disease progression (new or increases in size of previously documented lesions on imaging, clinical worsening, or rising biomarker concentrations). In the WHO 2014 classification, the definition for primary peritoneal cancers was revised. The outcomes review committee chair (NS) reviewed all 41 cancers previously classified as primary peritoneal as per WHO 2003 classification.

The outcomes review committee re-staged all ovarian and tubal cancers using FIGO 2014 criteria (previously staged using FIGO 2003) diagnosed in 2001–14.

Outcomes

The primary outcome was death due to ovarian (ICD-10 C56) or tubal (ICD-10 C57·0) cancer. Ovarian cancer includes primary non-epithelial ovarian cancer, borderline epithelial ovarian cancer, and invasive epithelial ovarian cancer. As stated above, ovarian cancer was defined using the revised WHO 2014 definition.

The site assignment is in contrast to the previous mortality analysis (censorship Dec 31, 2014), which used the WHO 2003 definition.
The majority (40 of 41) of previously classified primary peritoneal cancers using WHO 2003 criteria were reclassified as ovarian or tubal cancers. Secondary outcomes were ovarian and tubal cancer incidence and stage. For all outcomes, subgroup analysis was undertaken for invasive epithelial ovarian and tubal cancer. All outcome data were kept confidential until unmasking. Case fatality rate by stage was a post-hoc outcome.

Statistical analysis

At previous analysis (censorship Dec 31, 2014), there were 358 ovarian and tubal cancer deaths in the no screening group.

Compared with the no screening group, the mean estimated relative mortality reduction in deaths was 11% (Cox model p=0·24) in the MMS group and 9% (Cox model p=0·32) in the USS group. Any mortality reduction was only apparent about 7 years after randomisation. 162 (45%) of 358 of the deaths in the no screening group during 2001–14 occurred before 7 years. In 2015, for the no screening versus MMS or USS comparisons, we estimated that an additional 233 no screening group events would give 80% power at a two-sided 5% significance level for a difference in relative mortality of 25% during long-term (2015–20) follow-up, conditional on the observed mortality reduction of 11%. This estimate translated to a target sample size of 591 overall events in the no screening group: all 233 new and 73% (431 of 591) of total no screening group events would occur beyond 7 years. No formal adjustment was made to the test for having previously analysed the data in 2015 or making two screen group comparisons. Instead, we decided to openly describe the multiplicity issues and acknowledge the unadjusted p values. As the number of events were less than anticipated on the planned censorship date of Dec 31, 2018, follow-up was extended with a new censorship date of June 30, 2020.

Descriptive statistics regarding ovarian cancer death and incidence were created including tabulations of histology, stage, and screen type by group. The primary mortality analysis was changed from the 2016 report, in which we used a Cox version of the log-rank test,which is most powerful under proportional hazards. For the current analysis, we extensively discussed the best approach within trial management and trial steering committees, and consulted 12 independent international statistical, trial, and screening experts. The details and rationale underpinning this important change are reported separately. In short, given the accumulating external evidence of delayed mortality effects in screening trials, most experts supported the change in primary analysis to a test that was sensitive to delayed effects. We chose the versatile test that was agnostic to the specific form of the screening effect. The versatile test, described in 2016, is a combination test of three log-rank test statistics (Z1, Z2, Z3), covering early, constant, and late effects respectively (appendix p 3).
All analyses were by intention to screen. The primary mortality analysis was an MMS versus no screening and USS versus no screening analysis of the primary outcome using the versatile test, with a Royston-Parmar model to estimate survival differences. We defined survival time from date of randomisation (t0=0) to date of death due to ovarian or tubal cancer or censorship, or sooner if the volunteer died of another cause or was lost to follow-up. No allowances were made for screening non-compliance (study groups) or contamination (no screening group). We describe potential time-dependent features of the screening effect by estimating the hazard ratio (HR) and the absolute survival difference at the prespecified timepoints of 5, 10, 15, and 18 years (maximal follow-up was 19·3 years) using a flexible parametric Royston-Parmar model (appendix p 3).
We undertook two secondary analyses of the primary outcome. We fitted a proportional hazards Cox model to the primary outcome data. To allow for a formal analysis of the late effects of ovarian cancer, not subject to issues of data re-use and multiple testing, we also fitted a Cox model to the new data acquired since Jan 1, 2015. Both the methods and results of sensitivity analyses are detailed in the appendix (pp 8–9). Survival from diagnosis in women with ovarian and tubal cancer in the no screening group was also compared to national age and period adjusted survival rates at 1, 5, and 10 years. We undertook a subgroup analysis using the versatile test of invasive epithelial ovarian and tubal cancer death, in which other ovarian cancers were censored at death.
For the secondary outcome, cumulative incidence of ovarian and tubal cancer were presented graphically using standard Kaplan-Meier methods, based on time from randomisation to diagnosis. Death from other causes and bilateral salpingo-oophorectomy were censoring events. Administrative censorship was the same as for the mortality analysis (June 30, 2020). Ovarian and tubal cancer incidence rates were explored parametrically using a Royston-Parmar model that specifically allowed exploration of the underlying hazard functions for cancer incidence. For the secondary outcome of ovarian and tubal cancer incidence by stage, and the subgroup analysis of invasive epithelial ovarian and tubal cancers, we used incidence rate ratios with 95% CIs to compare no screening versus MMS and USS groups separately. We also calculated stage-specific ovarian and tubal cancer case fatality rates.
We used Stata, version 16 (versatile test verswlr function) for all statistical analyses. Results were independently verified by a different statistician using R, version 4.0.2 (versatile test logrank.maxtest of nph package). This trial is registered with ISRCTN, 22488978, and ClinicalTrials.gov, NCT00058032.

Role of the funding source

The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

Between April 17, 2001, and Sept 29, 2005, we invited 1 243 282 women to participate and randomly assigned 202 638 (16·3% of 1 243 282; figure 1).

We followed-up participants until June 30, 2020. The final cohort eligible for analysis consisted of 202 562 (>99·9%) of 202 638 women: 50 625 (>99·9%) in the MMS group, 50 623 (>99·9%) in the USS group, and 101 314 (>99·9%) in the no screening group. We excluded 76 (<0·1%) women (15 [<0·1%] in the MMS group; 16 [<0·1%] in the USS group; and 45 [<0·1%] in the no screening group; figure 1) as they had bilateral salpingo-oophorectomy, ovarian cancer before joining the trial, or had exited the registry before randomisation. As previously reported, baseline characteristics were balanced between the groups.

Screening ended on Dec 31, 2011. We undertook 673 345 annual screens: 345 570 in the MMS group and 327 775 in the USS group. Compliance with screening was high (81% in the MMS group and 78% in the USS group) with women undergoing a median of eight annual screens.