Strengths and weaknesses

We invited women from participating local health authorities’ registers. This is an approach often used in the United Kingdom.11 12 Other databases in use include electoral rolls, driving licence records, and commercially available directories.13 The alternative strategy is to advertise the trial extensively and let participants self refer.14 Recently, this has involved Google adverts and links to studies on a variety of websites.15 Women who volunteer to participate in research are often more educated and informed than the general population.16 17 18 Hence, invitation using health authority lists or electoral rolls is thought to result in participants who are more representative of the general population than those who self refer through adverts. Use of health authority lists provides additional logistical advantages. Recruitment rates can be controlled by varying the volume and frequency of mailing invitations, and electronic data transfer from health authority files leads to decreased data entry and improved accuracy. This is particularly notable with regard to NHS numbers, which are essential for cost effective follow-up of women through the UK Office for National Statistics. In a previous trial, in which women self referred, most women did not provide an NHS number.19 Finally, mass mailing allows acceptance rates and their variation between centres to be determined.

Recruitment rates vary with time and between centres. Maintaining this at target levels requires constant monitoring and individual solutions, as problems differ between centres. Support from the relevant patients’ support groups is crucial. For trials with lower acceptance rates, one solution is to increase the volume and frequency of mailing of invitations and local media coverage. Mass broadcast (television and radio) and print media are “low effort, high yield” recruitment strategies.20 21 In the UKCTOCS, as participation was limited to invited women, we were constantly trying to find the balance between too much and too little publicity. To this end, we avoided posters and flyers in general practitioners’ surgeries, well women clinics, and so on. Such publicity may have persuaded more women to take up their invitation.

In FLEXISIG, the sigmoidoscopy screening trial, the involved general practices checked the local authority lists and excluded 2% of 375 744 men and women as they were deemed ineligible owing to colorectal cancer, terminal disease, and so on.22 In the UKCTOCS, the 3266 general practices did no cleaning of lists, as we were concerned that this would affect timely mailing of invitations, which was crucial to maintain target recruitment. Thirty two of 1.2 million women invited complained about being contacted. This included four women with ovarian cancer out of an estimated 587 who would have been invited on the basis of national incidences.23 Given the small numbers excluded in FLEXISIG, the low rate of complaints in the UKCTOCS, and the substantial effort required, “cleaning up” of local authority lists does not seem to be necessary.

The overall acceptance rate was one in four (23%). However, the rates varied between different parts of the country. The data suggest that in cities such as London, Manchester, and Belfast one should expect lower acceptance rates in the range of one in five. However, in other parts of the country, such as Nottingham, Portsmouth, Bristol, and Derby, almost one in three women in the 50-74 age group might accept an invitation to participate in a screening trial. The varying uptake highlights the importance of running pilots at multiple centres. The UKCTOCS data are being analysed to see whether acceptance rates can be estimated by using age and the postcode derived index for multiple deprivation.

We could not account for undelivered invitations in our trial, as the initial invitations were posted in envelopes with no return address. In the recent FLEXISIG trial, 4% of invites were undelivered,12 and this is useful in calculating accurate acceptance rates. The overall acceptance rate in the UKCTOCS was similar to that seen in the colorectal cancer screening trial at Dundee and the WISDOM trial involving estrogen use after the menopause.24 25 These rates were substantially higher than the 4.3% acceptance rate reported after a mass mailing of more than 3.4 million in the US systolic hypertension (SHEP) trial.13 However, they were much lower than the 55% achieved in the UK FLEXISIG trial. One reason is the way randomisation was approached. In the UKCTOCS, we asked women to help to test a new screening programme for ovarian cancer. We told them that we would need to involve 200 000 women, half of whom would be screened and half would have the usual medical care. In FLEXISIG, participants were asked, “If you were invited to have the bowel cancer screening test, would you take up the offer?” The acceptance rate was based on those who answered “yes.”22 This emphasises the importance of wording of the invitations, an aspect sometimes overlooked in clinical trials. Meta-analysis of breast screening trials shows that use of direct contact strategies involving telephone or any type of personal contact can also significantly increase uptake. However, this is only feasible in smaller trials, and concerns about the cost effectiveness have been expressed.26 Another way to improve acceptance rates in screening trials is to invite people who regularly attend established screening programmes such as breast and cervical cancer screening, as they are self selected for their belief in screening.27 In the Million Women Study, 71% of women having breast screening returned the study questionnaire compared with 53% of all those invited.28 This was not possible in the UKCTOCS, as these programmes are limited to women aged below 65 in the UK.

In the UKCTOCS, 202 638 women were randomised in 4.3 years, which translates to a randomisation rate of 47 125 per year. The only trial to report higher rates was the FLEXISIG trial, with a randomisation rate of 68 173 per year.12 This was achieved by adopting a two stage recruitment design; 170 483 people who agreed to attend bowel cancer screening if invited were randomised, and only 40 674 allocated to the screen arm were recruited.29 This design, however, does mean that only limited data and no biological samples are available in the control group. In addition, recent revisions of laws such as the Data Protection Act and the Mental Capacity Act might lead to difficulties in follow-up through the Office for National Statistics and use of data in secondary studies. None the less, postal recruitment with attendance at the clinic only for people randomised to the study arm is an attractive design with significant savings in terms of time and resources and is definitely worth exploring.

Sophisticated web based trial management systems are becoming the norm for multicentre clinical trials.29 30 They enable simultaneous data entry from multiple sites by using standard web browsers with centralised data processing. In large multicentre screening trials such as the UKCTOCS, such technology ensures strict adherence to the protocol over a long period of time. Such solutions can be prohibitively expensive, however, especially for individual research groups. A move to encourage research and funding organisations to combine efforts to produce an open source solution for management of trial data is now occurring. This would empower a wider variety of people to do trials and may encourage more investigators in resource poor settings to take part in high standard research.31

A novel feature of the UKCTOCS was the use of an information video and group discussion during recruitment. The video ensured that all participants received high quality standardised information which was sustainable. Research nurses recruiting an average of 100 women a week find it difficult to deliver the same high quality message repeatedly. The video also helped with retention of experienced staff by relieving the monotony of repeating the same message many times a day over four years of recruitment. The group discussions often generated queries that helped volunteers to arrive at a more educated and informed decision than might have been possible if they were only seen individually in a busy recruitment clinic.

One in 6.5 women aged 50-74 in England, Wales, and Northern Ireland (total population 8 031 463) were invited to participate in the UKCTOCS.32 This is a major strength, as it makes extrapolation of the findings relevant to people planning randomised controlled trials in the UK. However, the population is not wholly representative of the UK, as the pre-identified Scottish centres were unable to participate.

Implications of the study

Running large multicentre trials is challenging. Senior investigators need to set aside time to support day to day trial management. Centralisation and automation of trial processes by use of web based trial management systems with high security encryption are essential. Information videos and group discussions allow sustained delivery of high quality standardised information during prolonged recruitment. The over-riding approach needs to be one that incorporates proactive management, flexibility, and individualised solutions. Close cooperation and regular communication between the coordinating and trial centre teams are key to success.