Explanatory or pragmatic

Chapter 3 began by reviewing the traditional classification of trials from phase I to IV. The fact that a promising treatment will rarely pass through just one trial in each phase was pointed out. This is particularly true of the phase III setting, where it may well be appropriate to consider a division of the phase III randomized trials into 'earlyV'developmental' trials and 'late'/'public health' trials. The former are designed in a manner which gives the treatment under investigation the best possible chance of showing its potential. This may mean testing new treatments in situations which do not reflect the nature or breadth of routine clinical practice. Only if they succeed in this setting are they likely to pass through to the latter stage, in which their effect in routine clinical practice is assessed. As stated in Chapter 3, this distinction almost entirely mirrors a classification of clinical trials introduced by Schwartz and Lellouch [6] - they referred to trials being either explanatory or pragmatic, the former having the aim of determining the precise action of a treatment in a defined setting, the latter answering the question - is the treatment of any practical use?

This is a somewhat idealized division, and it is not essential that all new therapies are evaluated in this way, but it can provide a useful framework for considering many design aspects of a randomized trial. In fact, the somewhat different aims of explanatory and pragmatic trials may affect the clinicians involved, the patients involved, the type of ancilliary care that patients receive, the type and amount of data that are collected, even the way the trial is analysed. These issues are summarized in Box 4.3, and described in more detail below. The basic principle is that the different types of trials value variability differently. In the early trial, which will generally be of modest size, it is a source of random error, which can limit the ability of a trial of modest size to detect a true underlying difference between treatments. All aspects of the design therefore aim to limit the sources of variability. When the issues turn to those of practical benefit of a treatment in clinical practice, heterogeneity can be a good thing - it can be explored in

Box 4.3

Two approaches to trial design




Determine precise action of a

Is the treatment of any practical

treatment in a defined setting



Homogeneous - those most

Heterogeneous - representative

likely to respond as we would

of those to whom the results


would be extrapolated


Those with most experience,

Representative of those who

most likely to get the 'best' out of

would use the treatment outside

a new treatment

of the trial


All aspects of patient care clearly

All other aspects of patient care


as for normal practice


As close as possible to the time at

At the time when treatment

which treatments diverge

policy would be decided


May require specific specialized

Only investigations used in

tests - intensive surveillance

routine clinical practice


Large amounts to identify

Minimum required to answer

protocol deviations and monitor

question posed by trial



May concentrate on those eligible on blind review

Analyse all patients

order to identify any clear evidence of qualitative or quantitative treatment differences and in the absence of this, confirm that the overall treatment effect is the best estimate of the effect in any individual patient.

0 0

Post a comment