Randomised Clinical Trials. David Machin
are to be the participants, then a definition of ‘healthy’ is required. This definition may be relatively brief or very complex depending on the situation.
At the early stages of the design process, one may have only a general idea of the types of subjects required although identifying the research question may have already made this reasonably clear. Thus, elderly patients may be the first target group for whom preventative action to reduce hip fractures may be considered. Then, when considering the trial question in detail, it might be decided to confine the elderly patient group to those residents in a nursing home. Further refinement may then, for example, define the elderly for trial purposes as those over 80 years of age and exclude those nursing homes who deal with psychiatric residents only. Considerations here might have been an anticipated very low fracture rate in those under 80 years of age and the difficulties associated with obtaining fully informed consent from patients with psychiatric illnesses. These selection criteria are easy to understand, easy to determine and therefore easy to apply in practice.
Eligible patients for the trial conducted of Example 1.7 by Meggitt, Gray and Reynolds (2006) in patients with atopic eczema had to satisfy an extensive list of criteria before they could be considered eligible for the trial. The general requirements specified patients 16–65 years of age with atopic eczema. However, only those with moderate‐to‐severe disease were to be included and this had to be determined according to the UK modification of the Hanifin and Rajka diagnostic criteria suggested by Williams, Burney, Hay, et al. (1994) which involves a detailed examination of the patient. Further, the trial excluded those: admitted to hospital for eczema; had used phototherapy or sunbeds; had been treated with cyclosporin, systemic steroids, Chinese herbal medicine, topical tacrolimus or evening primrose oil during the preceding 3 months; unstable or infected eczema in the previous 2 weeks requiring either highly potent topical steroids or oral antibiotics; TMPT activity <2.5 nmol/h/ml RBC; malignant disease: serious or uncontrolled systemic disease; HIV; pregnancy; lactation; mild eczema; and concomitant drugs known to interact with azathioprine. The process of checking the eligibility criteria excluded one‐quarter (24) from the 101 patients initially screened for entry to the trial although it is not clear how many of these were excluded on the grounds of having mild eczema as opposed to excluded for one or more of the other reasons. Clearly, if the mild cases are not to be included, the disease status should be determined early in the patient selection process to avoid unnecessary details from a patient who will not be eligible for the trial on that basis alone. One might also conjecture as to why those over 65 years are not included.
In general terms, it is not advisable to restrict the patient pool for entry into a clinical trial as if too many restrictions are in place, the patient pool reduces accordingly so that the required number of patients for the trial takes longer to identify and so trial duration may become prolonged. Also, if those that are eligible comprise only a small subgroup of the total patient pool, for example, selecting only the moderate‐to‐severe cases from all those with atopic eczema, then the trial results will not have application to those with less severe disease. As a result, clinical teams are left wondering whether the trial findings are applicable to these patients. Of course, in this example extending the eligibility to less severe cases might not be sensible if useful treatments are already available for these patients.
In determining eligibility, investigators should give as much thought as possible to ensuring generalisability of the subsequent trial results and therefore to define simple and minimal eligibility and exclusion criteria that allow the widest range of patients into the trial in whom benefit from the therapy may be anticipated. In many cases, these will include all the patient types for which the comparator in the trial is the current standard. If the eligibility criteria are too narrow then, however good the test treatment, the clinical implications will only be relevant to those small groups.
In the context of trials in oncology, but equally applicable in all areas, Wright, Bouma, Dayes, et al. (2006, p. 844) warn:
In a highly selected trial population, the question of generalizability becomes: how useful are the results of such a study in a more typical population of patients? Unfortunately such information is rarely reported, and oncologists and patients are left estimating how selected the trial population is and what implications it may have for applying the results of the trial in practice.
A long list of eligibility criteria is also very time‐consuming to produce and to verify. Nevertheless, and despite the requirement for as few restrictions as possible on the entry requirements to a trial, patient safety is of paramount importance and every care must be taken to ensure vulnerable patients are not entered into a trial. Those that are vulnerable, in the sense intended here, are for example those who may be at high risk of a serious adverse event were they to take the trial medication or undergo the trial procedures. This is also often a major concern of the committees giving ethical approval for trials to be conducted.
There must be specific reasons given for why a patient should not be included. For example, in some circumstances, pregnant women and those lactating (otherwise eligible) may be excluded for fear of impacting adversely on the foetus or the new‐born child.
One aspect of selecting patients for inclusion to a trial is seldom explicitly reported in the literature. This is the need to verify that each individual considered for recruitment is suitable for all the treatments or interventions on offer within the clinical trial. Thus, if there are three treatment options A, B and C, then not only must the attending physician be happy to prescribe every one of the options but also the patient must be willing to receive any of these options. If only two of these three are acceptable (by either the physician, patient or both), then the patient should not be regarded as eligible for the trial and so should not be regarded as a potential recruit. Further, if the physician thinks that for a particular patient one of the options is preferable then, despite eligibility in all other respects, the patient should receive that option and so again should no longer be considered for the trial. In such circumstances, the clinician should not enter the patient in the hope that following the treatment assignment process, the patient will be allocated the ‘preferred’ option. Neither should the physician expect, if the unacceptable alternative option is allocated by the randomisation process, then he or she can then simply withdraw the patient from the trial. Such action, certainly if repeated sufficiently often, will seriously undermine the trial and thereby a misleading outcome could arise. The difficulty is that we do not know to what extent the information or lack of information from such patients distorts or biases the estimate of the treatment effect that is obtained at the end of the trial.
Example 2.2 Trial eligibility – partial thickness burns
In a randomised trial by Ang, Lee, Gan, et al. (2001), patients with severe burns were emergency admissions into the specialist burns centre in Singapore requiring immediate treatment. Once admitted to the burns centre, only those patients with partial thickness thermal burns were eligible for the trial. Their consent was then sought and, once given, randomisation was accomplished by telephone to the statistical centre. Nevertheless, in certain cases, the attending medical team felt that the option of conventional therapy was more appropriate than the test therapy. For these patients, details of the clinical trial were not explained and conventional therapy commenced immediately.
2.4 The consent process
Once a potential participant is deemed eligible for the trial, fully informed consent is required before the individual can be formally registered for the trial and the intervention allocation process implemented. Of course, before consent can be sought those responsible for obtaining consent should regard the trial they are advocating as ethical from their own perspective. A simple ethical test that works for most researchers is: if your mother/father/child had this condition, would you be willing to enter them into the proposed trial? If not, are you sure the trial is ethical? Only if the consent seeker is in the state of equipoise, that is, has an indifferent opinion about the relative merits of the alternative treatments, is randomisation considered ethical.