Experimentation May Be Inappropriate

Trial size and rare events.

As osteoarthritis is so common, patient availability is not considered a problem for this trial.

However, the size of the this early phase II trial may prove to be of concern: small intervention groups may put the comparability of prognostic factors at baseline at risk, since the groups may be too small to evenly distribute the prognostic variables across all study groups, or to allow the identification of any adverse effects caused by the trial interventions (Koes and Hoving, 1998). The data analyses in the trial should be able to provide detail for the former but not necessarily for the latter. However, since the trial intervention includes established treatment components, such as exercise, it is believed that the risk of new adverse events occurring is unlikely.

Long term outcomes.

Randomised clinical trials are rarely able to evaluate long term outcomes of medical treatments even when the outcomes of interest reach well into the future (Black, 1996). Black aptly uses arthroplasty to illustrate this point. In this trial, follow up will continue to one year, the time by which most improvement after knee arthroplasty is seen with most improvement occurring within the first 3-6 months (Jones et al., 2007).

The influence of randomisation.

A randomised trial may also be inappropriate where thevery act of random allocation itself significantly reduces the effectiveness ofthe intervention (Black, 1996). Black’s commentary argues that this can arise when the effectiveness of theintervention is dependent upon the active participation of the subject which then depends on the subject's beliefs and preferences.

This can result in a lack of any subsequent difference in outcomebetween comparison groups

thereby risking an underestimation of the benefits ofthe intervention. This seems to be a similar issue, relating to potential bias, as that discussed in the earlier section of placebo/control groups; the same argument, that any subsequent bias is not considered detrimental since it would reflect clinical practice, could be forwarded again in rebuttal.

Effects of Randomisation.

There is evidence to suggest the belief that patients cite their aversion to randomisation as a main reason for their refusal to enter a trial (Llewellyn-Thomas et al., 1991) which is of concern to all researchers seeking as representative a sample as possible. It was hoped that, because the proposed trial aimed to assess an additional intervention on top of usual care (with no placebo/control group) that this would not prove to be a serious issue. Randomisation is controversial; the feasibility of performing randomised clinical trials, the appropriateness of the design, ethical issues and the overall value of randomised trials in obtaining generalisable results are all areas of debate (Abel and Koch, 1999). This debate is not helped by randomisation being imbued with powers it does not necessarily possess; the erroneous belief that a researcher ensures comparability between treatment groups merely via randomisation thus ignoring differences occurring between groups following randomisation and before outcome / response measurement, which can adversely affect this comparability (Abel and Koch, 1999). Treatment effects, random errors, and patient / clinician motivation are amongst factors that Abel and Koch list as causes for treatment group differences. Also a dilution of intervention effect may be caused when prognostically heterogenous, rather than homogenous, groups occur after randomisation (Koes and Hoving, 1998). Plus, randomisation is not the sole basis for tests of statistical significance or for causal inferences on treatment effects, or sufficient to guarantee blinding (Abel and Koch, 1999).

Ethical Issues.

Ethical issues may also make a trial impossible to conduct. McLean (1995) states that there is a general acceptance that a study which is not scientifically valid is unethical; such studies are an abuse to the patient group concerned and a waste of scarce research resources. The need to obtain ethical consent prior to performing a study, from a relevant Ethics Committee, is seen as a way to assess the value of the study in these terms and to protect the participants in studies from unacceptable harm. Often, trials which expose patients to an intervention which is believed to be inferior to current treatment are regarded as unethical (Sibbald and Roland, 1998).

In a more subtle issue, Lilford and Jackson (1995) raise the question of whether or not trials are unethical without the prerequisite of equipoise. Equipoise is defined as:

“the point where there is no preference between treatments, i.e. it is thought equally likely that treatment A or B will turn out to be superior”

Lilford and Jackson (1995)

Lilford and Jackson (1995) argue that, whilst for many trials equipoise is desirable, there are times when equipoise is unachievable, an example might be when a central authority decides that a treatment needs evaluation for reasons of public interest. They also argue that randomisation without equipoise may actually be desirable in some instances, for example where access to an intervention is limited through scarce resources.

Contamination.

Black (1996) also describes the possibility of contamination if clinicians involved in a trial have to provide more than one intervention arm since the way in which the arms are delivered to patients may be influenced from this. Whilst recognised as an issue the current trial was designed in such a way that this could not happen so was not a concern for this trial.

Number and Complexity of Interventions.

As a further issue, health service interventions in primary care are often far more complex than the interventions assessed in clinical trials (Mant, 1999; Black 1996). Assessing very complex interventions can make trials impossible (Black, 1996). And, even when a trial is possible, the generalisation of findings to patients with their own individual complexities is difficult (Mant, 1999). The integration of qualitative research into the randomised trial can assist understanding of complex interventions (Mant, 1999) but cannot fully address this limitation. The Medical Research Council’s Framework for the Development and Evaluation of RCTs for Complex Interventions to Improve Health (MRC, 2000) was developed to assist trials evaluating complex interventions (see next section on the development of the trial intervention) but cannot fully address the problem.