Randomised Clinical Trials. David Machin
to determine the relevant trial endpoint(s). Also, aspects associated with analysis, reporting and interpretation of the results are included. Finally, we introduce the basic ideas of a statistical model upon which the ultimate analysis of the clinical trial is based.
2.1 Introduction
Although in this chapter we will focus on one particular design, there are many features of the clinical trials process which are common to the majority of situations. Thus, we use the example of a parallel two‐group individually randomised trial to overview some pertinent issues, ranging from defining carefully the research question posed, and thereby the type of subjects to recruit, the interventions used, the allocation of the trial participants to these interventions, endpoint assessment, analysis and reporting. However, in later chapters we will expand on detail and also on other design options.
This basic design will often compare a Test therapy or intervention with a Standard (or control) therapy. Frequently, the patients will be assigned at random to the options on a 1 : 1 basis. In reality, the actual choice of design will be a key issue at the planning stage and it should not be assumed that the common design used here for illustration best suites all purposes.
Example 2.1 Two‐group parallel design – symptomatic oral lichen planus
Poon, Tin, Kim, et al. (2006) describe a randomised trial of the comparative effect of the then current Standard, topical steroid (S), and the Test, topical cyclosporine (C), in patients with systematic oral lichen planus. The basic structure of their trial is given in Figure 2.1.
This trial typifies the design and structure of the randomised parallel‐group comparative trial which is used extensively. The key features include the following: defining the general types of subjects to be studied, identifying the particular subjects eligible for the trial and obtaining their consent, randomly allocating the Standard and Test interventions to individuals and, once the intervention is introduced, making the appropriate assessments in order to determine outcome. The analysis of the data recorded for all patients will form the basis of the comparison between the intervention groups and (hopefully) provide a clear indication of their relative clinical importance and supply the framework for the subsequent clinical report describing the results.
In the following sections, we will follow the sequence of Figure 2.1 even though this will not be entirely reflected in real‐life situations of the planning stages where a back and to process is more likely. In general, there will also be additional and trial‐specific steps that one must also make.
Figure 2.1 Individually randomised controlled trial of the effect of topical steroid (S) and topical cyclosporine (C) in patients with symptomatic oral lichen planus.
Source: Based on Poon, Goh, Kim, et al., (2006).
2.2 The research question
Of fundamental importance before embarking on a clinical trial is to identify the research question(s) of interest. Such questions may range from a very scientifically orientated objective to one focussed on a very practical day‐to‐day clinical situation. For instance, the trial of Example 1.11 aims to reduce the number of hip fractures in elderly residents of nursing homes, whereas one objective of those reported in Example 1.1 is concerned with gastric emptying times and is similar to a non‐clinical laboratory‐based investigation.
Two key issues are as follows: Is the question worth answering? Is the answer already known? Clearly, the answer to the first should be unequivocally a ‘Yes’. For the second, one might expect a ‘No’, although there are circumstances when an earlier result may need confirmation. For example, Wee, Tan, Tai, et al. (2005) conducted a confirmatory trial of one previously conducted by Al‐Sarraf, Pajak, Cooper, et al. (1990). The rationale for the repeat trial was based on the former trial involving mainly Caucasian patients, whilst the repeat was to involve those of predominantly Chinese ethnicity and to be conducted by different investigators in another part of the world. In the event, the advantage to chemo‐radiation as compared to radiotherapy alone in terms of overall survival of patients with advanced nasopharyngeal cancer was confirmed thereby indicating wider generalisation of the results from the two trials.
The question(s) posed must have important consequences in that the answer should inform research and/or influence clinical practice in a meaningful way. Further, before the trial is conducted, there should be a reasonable expectation that the trial question will be answered otherwise, for example, patients may be subjected to unnecessary investigation and possibly discomfort without justification. One exception to this condition may be when considering information from necessarily small randomised trials in truly rare diseases or conditions where patient numbers will be insufficient for the usual rules for trial size determination to be applied. We return to this latter issue in Chapter 20.
Erbel, Di Mario, Bartunek, et al. (2007), see Example 1.12, state in the summary of their trial that:
Coronary stents improve immediate and late results of balloon angioplasty by tacking dissections and preventing wall recoil. These goals are achieved within weeks after angioplasty, but with current technology stents permanently remain in the artery, with many limitations including the need for long‐term antiplatelet treatment to avoid thrombosis.
This provides a clear rationale for testing a stent for coronary scaffolding made from a bioabsorbable material which should provide an effective scaffold but would not be permanently retained in the artery.
Meyer, Warnke, Bender and Mülhauser (2003), Example 1.11, point out that hip fractures in the elderly are a major cause of disability and functional impairment so that reducing their incidence by encouraging the use of hip protectors within nursing homes may help to reduce this morbidity.
The use of homoeopathic remedies is widespread although there have been few randomised trials to establish their relative efficacy against conventional methods (including placebo). Consequently, the trial of Example 1.4 by Stevinson, Devaraj, Fountain‐Barber, et al. (2003) compares daily homoeopathic arnica against placebo in patients with carpal tunnel syndrome having elective surgery, as it has been claimed that:
Homeopathic arnica is widely believed to control bruising, reduce swelling and promote recovery after local trauma: ….
At least to the investigators and also the corresponding journal editors and peer reviewers, these trials address important questions. The results of the trials suggest that the bioabsorbable magnesium stent is a useful development, encouraging the use of hip protectors reduces hip fracture rates, whereas homoeopathic arnica appears to be no better than placebo.
2.3 Patient selection
Common to all clinical trials is the necessity to define precisely which types of subjects are eligible for recruitment purposes. This implies