The Cancer Journal - Volume 11, Number 4 (July-August 1998)
An integrated methodology for therapeutic trials combining rigor and audacity
Clinical research has now become extremely sophisticated, particularly with respect to clinical trials and especially those designed to test the value of new drugs, that it could be considered as a model of scientific rigour applied to medicine. Results obtained within such a rigid methodological framework, including the regulatory, legal and ethical aspects, appear to be irrefutable. Is this not medicine based on hard facts?
However, how many of these new drugs which cause such a stir on the stock market are simply "me-too", without bringing any real progress, and how many others prove to be less effective or more difficult to use than was initially claimed by their promoters, in spite of the rigorous methodology employed?. Let me remind the reader of the tenets of this methodology:
1) Every clinical experiment should be designed to answer a single question; that is, one starting hypothesis of the type: is the new molecule presented as drug A more effective than the placebo or than the established drug B in a particular indication, a well defined disease at a given stage of its evolution, with predetermined criteria of evaluation. We could take as an example pre-menopausal women operated on for breast cancer, with demonstrated invasion of regional lymph nodes but without detectable distant metastases, who would be evaluated by the frequency of and time to relapse, and by survival time.
2) The experimental and control groups must be randomised, to take into account the inevitable variability within the patient population.
3) Whenever possible, treatment should be performed according to the double-blind method.
4) Biomathematicians should use appropriate statistical methods to ensure that the results are free of any identifiable experimental bias.
5) Finally, the trial must respect ethical guidelines and good clinical practice.
In the face of such strict criteria, is there any need for a critical look at these procedures? Of course there is, at least in this Journal. Without compromising their rigour, they could be adapted to encourage imagination and progress, and thus benefit patients more.
We all know that clinical trials have become extremely complicated and expensive because a large number of patients must be enrolled. As a result, not all possible therapeutic regimes can be tested and a severe selection process takes place. Financial considerations weigh heavily in this selection and, through collaborations with the pharmaceutical industry, market forces play an important role in the final choice of which trials are conducted.
Let us take the example of a population of patients in whom two drugs, A and B (the established reference), are to be compared. It is possible that the two drugs act by mechanisms which are different enough to divide the population into four sub-populations: those who respond to A (G1), those who respond to B (G2), those who respond to both A and B (G3) and those who do not respond to either A or B (G4). If there are many more patients who respond to A alone than to B, then G1 + G3 >> G2 + G3.
According to the classical protocol, the trial would conclude that A is more effective than B, because the sub-populations cannot be distinguished. The trial was not designed for that. The market targets the whole population and not an undetermined fraction thereof.
This idea is not new, but it serves as a example to show how, in a simple situation, the deliberate choice of formulating a single hypothesis, as in most clinical trials, means that part of the information generated is missed or not exploited.
The real situation is probably more complicated than the one described above. The heterogeneity among a patient population is very large, even when they have been selected according to precise criteria, which are often based on the results of "staging". It is true to say that each patient has a separate disease, if only in that their genetic make-up, personal history and associated diseases are different. During controlled clinical trials this heterogeneity is considered as an obstacle to discovery and to progress in therapy, which has to be overcome by randomisation and the use of large series. This is a lesson learnt from statisticians.
Here I am suggesting that we should use this heterogeneity and all the information defining it to improve our therapeutic arsenal. This is a complete reversal of the normal situation, in which only a small proportion of the available information is exploited, since the rest appears worthless. I propose "recycling" all the data which are usually ignored to obtain new scientific information. Of course, not all these data are exploitable but, by analysing the factors contributing to the heterogeneity of the population using modern mathematical methods (1), we can hope to achieve therapeutic innovation through the invention of new molecules. Cancer is an appropriate area in which to develop this research strategy. When we consider the currently established drugs, most of them are only indicated for one particular type of cancer. This suggests that comparative tests for new molecules should be carried out with subjects suffering from cancers of different organs and at different stages. I know how unorthodox this suggestion will seem. I am less sure whether I have succeeded in explaining the rationale behind it. I am ready to discuss it with any colleague who wishes to question the too-rigid approach of conventional clinical trials.
To contact the author...Click here Thank you.
1. Salomon J-C. La recherche sur le cancer dans l'après-génome. Cancer J. 11, 102-103, 1998.