Thursday, 21 December 2017

The randomised controlled trial

The randomised controlled trial is regarded by many as the most appropriate research design to evaluate the effectiveness of an intervention (Littlewood 2011).

Two features of a basic randomised controlled trial will be considered in this blog, and the clue is in the name – randomisation and control.

Firstly, in a randomised controlled trial, the research participants are allocated to two or more groups randomly, meaning by chance. Secondly, in its most basic form, research participants are allocated to the intervention or control group. Figure 1 offers a schematic representation of the most basic randomised controlled trial.

Figure 1 The randomised controlled trial

Taking the components of a randomised controlled trial in sequence, random allocation will be considered first. Imagine a physiotherapist who has designed a new exercise intervention for the treatment of low back pain and wants to establish the effectiveness of the treatment. She receives the first patient in the clinic; a young male with mild low back pain who continues to work and enjoys exercising. She decides that the patient should receive the new treatment and allocates him accordingly. The second patient of the day is also a young male but in contrast has not worked due to severe low back pain for a number of years and does not appreciate the benefits of exercise. She decides that this patient would probably not respond well to the new exercise intervention and so allocates him to the usual physiotherapy treatment. This non-random process of selection continues until all 100 research participants are recruited. Upon completion of the trial the physiotherapist notices that patients in the intervention group were more active and, on average, reported milder levels of low back pain. This means that the groups being compared are different to begin with, even before they are exposed to the treatment.

After the data is collected, the analysis suggests that the new exercise intervention is far superior to usual physiotherapy. Is this a valid conclusion? Probably not; because as physiotherapists we know that people with less severe pain, who maintain good levels of activity, tend to perform better than those with more severe pain and greater functional deficit, irrespective of the treatment they receive. This is an example of selection bias where the process of allocation of participants to groups has been adversely influenced. In contrast, a process of random allocation would serve to distribute participants between the group in a more balanced manner according to the factors we know, for example age, gender, condition severity; but importantly also the factors we don’t know, for example psychological status, attitude to treatment etc. This provides a more appropriate basis upon which to judge the effectiveness of an intervention because the only significant difference between the groups should be the treatment received rather than other factors that can influence the outcome of treatment.

I often reflect that in a fair race Usain Bolt would beat me over a 100-metre race, but give me a 80 metre start and I think I would cross the line first. Seems obvious when put in those terms, but we also need to apply this thinking to research; if the race isn’t fair, i.e. if one group is different to begin with, it is perhaps unsurprising that one treatment appears superior to another in skewed conditions.

The second component, the control group, will now be considered. It is easiest to recognise the value of a control group by considering a situation where it is not used. Another example; following a day’s lectures, my students often complain of headaches. 

So, I ask them to rate the severity of their headache on a scale of zero to ten where zero equates to no pain and ten equates to the worst pain imaginable. I then instruct them regarding an exercise intervention that I have spent many months developing; repeatedly elevating their left arm. I’m not sure why this intervention works but I’ve spent so much time thinking about it that it had to be worth something, surely?! I ask the students to repeat the exercise two or three times before returning the next morning. Consistently, I find that the severity of the students’ headaches had reduced significantly by the following morning. Should I conclude that my exercise intervention is effective based upon this data? Clearly not, it is much more likely that not having to listen to my voice, doing something more enjoyable, having a good rest and other factors are more likely to explain the reduction in headache severity. If I had randomly allocated students with headache to the intervention group, who undertook the exercise, or control group where students continued as usual without the exercise then a more valid evaluation would be possible. When the students returned the following morning we could establish the headache severity of both groups knowing that the only difference between the groups was that one received the intervention and the other didn’t because the control group would also be exposed to all these other factors, for example rest, that might be responsible for a reduction in headache severity. In this situation it is highly unlikely that my exercise intervention would prove its worth and hence the value of a control group is highlighted, and I need to go back to the drawing board…developing an intervention with a plausible mechanism of action which is based on current evidence.

An important point to recognise in relation to physiotherapy research is that the control group does not usually consist of no treatment; this is one of the most common misconceptions relating to a control group – so much so that I have a standard response paragraph for peer reviewers when they make this point. The reason for this is that it is generally accepted that it is unethical to withhold treatment of potential benefit from patients. So most randomised controlled trials involving physiotherapy compare the new intervention against a placebo, for example de-tuned ultrasound, or usual care, for example usual medical care. Hence the new intervention is expected to demonstrate that it is at least equivalent but usually superior to existing treatment before it would be considered for implementation into practice.

Hope this background information was useful; on the back of this my next post, hopefully, will be a critical appraisal of a randomised controlled trial I was involved with, the SELF study. Thanks for reading, would really appreciate your constructive thoughts.

(Adapted from UnderstandingPhysiotherapy Research by Littlewood & May)



No comments:

Post a Comment