Our survey collected valuable information on how the UKCRC registered CTUs monitor phase III randomised CTIMP trials. The main finding is the wide variety of ways in which central and on-site monitoring are conducted. The survey showed for central monitoring a variety of use, method, frequency of execution, method of trigger assessment and items in triggers. For on-site monitoring, variety was shown in who attended, how many attended, how long site visits were, how to decide when to visit a site, the determinants of a site visit and what to do when there, in terms of SDV and monitoring processes.
Although 36/37 (97%) of CTUs used central monitoring to guide, target or supplement on-site visits and 20/36 (56%) run central monitoring at least monthly, for 19/36 (53%) the central monitoring is not programmed (fully automated). This would likely form a large and repeated burden on the trial management team adding to the cost of running the trial. It should be possible to fully program this work, with the programming overhead being less than the trial management team burden for longer trials.
Our survey showed that a variety of people are involved in the on-site monitoring. Though this is related in part to the differing job titles used in the CTUs, it would be good to know if the 19/34 (56%) of CTUs using specific monitors found there was an advantage in this practice.
The variety of items triggering on-site monitoring and their varying frequency of use shows that there is considerable scope for prospective research to better specify where CTUs should expend their energy. Whitham et al (14) suggest 8 ‘performance metrics’ to be used for all trials alongside just one or two trial-specific metrics. TransCelerate has published 8 overlapping metrics(15). These suggested metrics need prospective test and reporting of experiential data to show whether they work.
Our survey showed that many CTUs claim to do 100% SDV of some data. However, the only publication looking at this has shown 100% SDV does not change the primary efficacy trial results (7). SDV may be necessary if particular data have been assessed as being critical to the trial and only able to be monitored in this way, but the FDA has set out guidelines (2) encouraging trialists to consider the needs of each specific trial rather than assume that all trials need SDV.
The Morrison survey from 2009 (9) acquired data from 65 mainly US groups performing trials and concluded that there was a wide variety of monitoring practices and called for research to develop an evidence base for monitoring practice. Our UK survey, nearly a decade later reaches the same conclusions. The previous surveys of UKCRC registered CTUs in 2011 (10) and 2017 (Elizabeth Swaby, personal communication) agreed with the current survey, also finding that all CTUs at least sometimes informed their monitoring plan through a risk assessment and that most CTUs used some level of central monitoring, either with or without on-site monitoring.
There are several limitations to our study. Although the response rate of 86% was high (and double that of the other monitoring surveys), there was complete data on only 34 CTUs. This may be considered a small number to represent UK clinical trials. It is limited to the UKCRC registered CTUs who have demonstrated a sufficiently high standard to achieve UKCRC accreditation. These are not fully representative of academic clinical trials units in the UK who are not registered, though fewer randomised Phase III CTIMP trials are run outside registered CTU. The monitoring set up of academic trials units in other countries and of industry-led or contract research organisation coordinated trials may differ. However, the information here should be useful to all groups running trials of any phase.
As survey data of policy across a CTU, this information on monitoring lacks detail and does not represent the monitoring that happens on a specific trial. This is more obvious on the SDV questions where 8 CTUs felt the SDV was so variable they could not complete the table. It may be that our results are of an ideal world situation.
Carrying out this survey has highlighted the need for clarity with regard to terminology in monitoring; multiple definitions for single terms are problematic when communicating with practitioners and researchers across the field of clinical trials. For example, in our survey we used the word ‘triggers’ to describe the items that are looked at to decide whether a site visit is needed. However, trialists and researchers use triggers, metrics and there may well be other names that our team has not yet come across. On this point, we favour using the word “metrics” and considering visiting a site if the metric threshold is breached. As a community, we need to show how the quality tolerance limits mentioned by ICH (1) relate to metrics. There is also confusion between data cleaning and monitoring, particularly since a monitoring outcome is to prompt for data corrections/clarification. Clarifying the terminology will go a long way to enabling us to move this area forward.
Some research on whether monitoring is best carried out by dedicated monitors, both centrally and on-site would be useful. Our terminology of jobs and some idea of the job description part of whoever is tasked with the monitoring would be beneficial.
With the strong steer from the surveyed CTUs, the next priority is optimising the central monitoring. We think the metrics that are used to improve the data integrity and patient safety need research to confirm that they are helpful, to select which are required, to clarify how they should be used (particularly in terms of frequency and post-metric actions) and to consider where they sit (are they a way of improving the data integrity and the patients safety or are they a way of judging it?).
We think there is research work to be conducted on how on-site visits are best performed. Should we be looking at SDV, processes or a mixture of these? And are site visits necessary or could the effect of a site visit be replicated in other ways by staying at the CTU?
We would like monitoring to get to the same place as protocols and statistical analysis plans for clinical trials have reached (Spirit guidelines,(16),Statistical analysis plan guidelines, (17)). Trial Forge (18) is aiming to increase the evidence base for efficient trial monitoring. We see a practical monitoring guidance document that shows, for differing risks of trials and individual processes, the monitoring that is required. This could perhaps be in the form of a template data monitoring plan.