Randomized controlled trials

Introduction

Randomized controlled trials (RCTs) are well established as the gold standard approach for making comparisons between interventions. ¹ Much methodological research into RCTs has been performed and a lot of guidance exists to ensure that they are performed to the highest standards. ^2,3 In these few paragraphs, we shall consider some of the basic concepts of an RCT that make it such a powerful research tool. We shall consider a parallel arm RCT with a superiority endpoint. Therefore, imagine we wish to compare a new intervention to a standard of care treatment. Each individual recruited to the study will be randomized either to receive the new intervention or to receive the standard of care treatment. We are interested in investigating whether the new intervention results in a better outcome.

Randomization

Randomization has been used in medical studies for over 60 years. ⁴ By allocating individuals to study arms purely by chance, we find that in the long run the characteristics of participants are similar between study arms. For example, in general we find that roughly the same proportions of women are in each study arm, the average age is similar between groups, and so on. This identifies one of the major strengths of randomization. We can see that without it we may find that we have, consciously or subconsciously, preferentially recruited a group of individuals with a better predicted outcome to one of the study arms and this could bias the results of our study. ⁵ The real power of randomization lies in the fact that this is not only true for known and measurable factors, but also for any unknown and unmeasurable factors.

It is generally accepted that in general we do not need to perform statistical tests on a baseline characteristics table. ⁶ This is because if randomization has been performed correctly, one would expect around 5% to be significant with P < 0.05 just by chance.

Control group

A second key component of an RCT is the presence of a control group. By entering into a clinical study, and perhaps receiving more intensive monitoring as a result, we may find that individuals would have had an improved outcome anyway regardless of the intervention administered (the so-called Hawthorn effect). ⁷ By including a control group that receives the standard of care option, we have an estimate of the response rate under these trial conditions in the absence of the intervention.

Blinding

A study is single blind if the study participant does not know which arm of the trial they have been allocated to. A double-blind study is one in which both the participant and the clinician/nurse and other health professionals are unaware of each individual's allocation. ^2,3 The reason for blinding is to ensure that individuals in both arms are treated equally. Thus, as we know (from randomization) that any differences in characteristics at baseline are due to chance and we have treated individuals similarly during follow-up, we know that the only systematic difference between study arms is the fact that one arm received the intervention and the other the control treatment. Therefore, any differences in outcome may be attributable to the intervention itself. Of course, it is not always possible to blind an RCT. One example may be if we are comparing the outcome of key-hole surgery to open surgery. However, clearly we should still make every effort to ensure that all individuals in the study are treated equally to minimize any potential biases.

Consolidated Standards of Reporting Trials statement and flow diagram

Despite the fact that RCTs are a common study design, there remains a wealth of evidence suggesting that the reporting of RCTs is still inadequate ⁵ (the inadequate reporting of one component of this, sample size calculations, was mentioned in the previous edition of Phlebology ⁸ ). The Consolidated Standards of Reporting Trials (CONSORT) statement, with guidelines most recently updated for 2010, represents an effort by leading international researchers to offer guidelines to manuscript authors regarding the essential items that should be included in RCT reports. ⁹ The CONSORT statement has been supported by more than 400 journals and several editorial groups, and is clearly an excellent source for guidance when writing up the results of an RCT. ⁵ One major component of this is the CONSORT flow diagram, which provides a clear summary of the disposition of study participants before and during the course of the RCT.

Disadvantages

As can be seen from above, RCTs are the gold standard approach for assessing the efficacy of interventions. However, like all study designs, RCTs have some limitations. Firstly, as they are costly and require intensive follow-up, they tend to only observe patients for short periods of time. Therefore, meaningful information about clinical outcomes and side-effects that can take decades to manifest are rarely available. Furthermore, a major limitation of RCTs lies in the type of study participants who are recruited to participate. It has been shown that those who choose not to participate or who are excluded from an RCT can have different characteristics to those who are included. ¹⁰ This can be as a result of the study inclusion/exclusion criteria, and/or as a result of the type of people who are likely to be willing to participate. Although these differences in the study population will not lead to biased comparisons, they may lead to an underestimate of the incidence rates present in the general population of interest. Finally, it is clearly unethical to randomize to some interventions. For example, it would be unethical to study the impact of excessive alcohol use on the occurrence of depression using an RCT.