Can randomised controlled trials be more efficient?

Well-conducted randomised controlled trials remain the most reliable way to demonstrate the true efficacy and cost-effectiveness for the majority of medical treatments. However, they are not without limitations. The average cost of conducting a randomised controlled trial can run into hundreds of thousands, usually millions, of pounds. It may also be several years before the results of the trial are made available to use.

It is therefore not unsurprising to see a greater call, particularly from funders, for study designs that support ways to make the conduct of a randomised controlled trial more efficient, without compromising reliability and thus introducing bias.

The recent “Efficient Study Designs” call by the National Institute for Health Research (NIHR), made just such a request. The call suggested features in which studies might incorporate greater efficiency. Examples included: the use of electronic health records (e.g. to identify patients who might participate in a study or for follow up); innovations to promote rapid or more efficient recruitment and follow-up; simplified design and/or follow up that focuses on a limited number of the most important outcomes and adaptive designs for trials.

Case study: use of electronic health records to increase efficiency in clinical trials

Electronic health records (EHRs) are a way of viewing a patient’s medical record via a computerised interface. They are routinely stored in health centres, e.g. general practices, across the UK, and every resident who registers with a GP usually adopts an EHR. There are established benefits and drawbacks from using EHRs, although one of the growing advantages is the ability to use anonymised EHR data to carry out “real-time” health research to deliver service improvements.

The argument for designing pragmatic randomised controlled trials that use routine electronic health records has been put forward, and a recent study in the BMJ Open provides some insight into how EHRs, when combined with a text messaging intervention, might increase the efficiency of outcome reporting within an RCT setting.

Many general practices use text messaging as a way of contacting their patients, indeed the NHS provides guidance on this. Systematic reviews have demonstrated some clinical benefits from text message use, for example in diabetes self-management, smoking cessation and medication adherence.   However, few, if any, studies have examined whether text messaging improves adherence to promotional campaigns that seek to increase the uptake to the seasonal flu vaccination, in those patients eligible for it. The TXT4FLUJAB study was designed as a cluster randomised trial and carried out over 156 UK general practices. The selected practices were familiar with text messaging use, but did not (at the time of the trial) use it to remind patients to come in for their flu vaccination.  Practices were randomised to either do what they would do normally to promote seasonal vaccination uptake (e.g. posters, letters to patients etc.) or were randomised to provide this standard activity plus send eligible patients a text message containing a reminder to come into the practice for their vaccination.

The use of text messaging led to a modest increase (absolute value 2.62% [95% CI −0.09% to 5.33%]) in the uptake of vaccination, the primary outcome. However, what was particularly interesting was the demonstration that most of the primary outcome data could be obtained from information that was routinely collected through the various IT systems already embedded within the GP practices. Such data was usually immediately coded into the patient’s electronic health record at the time of vaccination uptake, making it easier to search for and extract, thus saving research time. For these sorts of patients, the authors impressively reported no missing outcome data. There were some limitations with TXT4FLUJAB. For example, the authors point out that the proportion of patients that have a mobile telephone number registered with their GP tends to be low. They also noted contamination between the intervention and standard care arms in the numbers of text messages beings sent out, thus affecting the power of the study. Issues around consent were also discussed as limitations for wide spread immediate adoption. Nevertheless, in terms of efficiency, the authors demonstrated that their design was both feasible and cost effective, stating that the overall cost of the trial was under £100k which, given a reach over 100,000 patients, equated to a cost of less than £1 per patient recruited, very impressive for an RCT of this size.

The use of EHRs to provide clinical trial outcome measures has the potential for greater use and adoption. A number of other studies have demonstrated feasibility but have also highlighted potential challenges; such as the need to justify if the quality of the collected data is sufficiently high and complete to reduce any risk of introducing bias or confounding. Furthermore, electronic health records may not always collect data related to a specific outcome of interest, and are likely to be ethical issues around data sharing which will need to be overcome.

Nevertheless, methods that support the development of a reliable pragmatic RCT, with features of an efficient study design, can demonstrate, not just cost savings, but may also provide clinically useful information in a timelier manner. As a result, there are already broader discussions and initiatives to support this development, and I suspect we will see a lot more of these efficient methodologies in future clinical trials.

Disclaimer: The views expressed are those of the author and not necessarily of any of the institutions or organisations mentioned in the article.