Data monitoring committees in clinical trials: best practice, complexities and considerations

What is the problem all about? Clinical trials are sometimes described as being rather ‘blunt’ instruments, and that often may be true: we use them to find answers to very broad questions such as ‘does this drug work better than a control drug?’ But they are also very important and delicate instruments: we use them to find answers to questions such as ‘how much does this drug work?’ ‘what is the safety profile of this drug?’ and so on. And in the nature of research, we do not know what the right answer is and we have to rely on the clinical trial to give us the right answer. If anything is broken in the trial, it might give us the wrong answer – but we might have no way of know whether answer is right or wrong (unless we can see a break, or maybe just a dent, in the trial). Bias is one of the big concerns in any trial and it is why we typically use the key building blocks of blinding and randomization. It is also why we typically do not look at the data every week or two and see how one treatment is doing compared with another. We set up an experiment, we allow it to run unhindered, and then we look at the results. The problem (of course) is what should we do if something in the trial is going in a very different direction to where we thought it was? This might be that the new treatment is far better than the control treatment – even to the extent that it is far better than we initially expected. Patients are potentially being disadvantaged by being randomized to the control arm; something needs to be done. Or it might be that our new treatment, while backed by some of the greatest investors and optimists the world has produced, actually is not working. We are wasting a lot of time, resource, patients’ goodwill and, not to forget ... quite a lot of money. Something surely needs to be done! So we are in the dilemma of wanting to run a trial, without external interference, maintaining blinding and randomization, but we also want to know what the results look like. Enter the independent data monitoring committee (DMC).