Printed in Great Britain LEADING ARTICLE Tropical Trials and Tribulations

Field trials are the cornerstone in the evaluation of new public health interventions in the developing world. There is a tendency to base the introduction of new public health measures on very few trials. This is understandable given the expense in terms of money and time in mounting these trials and the enthusiasm to do something about the persistently high mortality in many parts of the world. However in the conduct of vaccine trials in West Africa a number of problems of design and analysis have presented themselves. These problems do not seem to be widely recognized but they affect the design of new trials and the interpretation of those already published. Some of these points are rele-vant to other forms of intervention study, such as vita-min A supplementation, and to trials in any part of the world. They fall into three broad areas; random-ization, speculative versus practical trials and outcome measurement. RANDOMIZATION There is no need, one would hope, to reiterate the arguments for randomization in field trials of interven-tions. If the randomization is at the individual level this presents few additional problems compared to clinical trials, except for size, unless the intervention has an effect on those who are not in the intervention group. For example, if an intervention against mosquitoes by the use of individual insecticide impregnated bednets reduced the total vectorial capacity in a village then some account would have to be taken of this protective effect on the control group. This is a situation in which randomization by social group or community may be appropriate. A much more common reason for adopt-ing this design is the logistic requirements of delivering the intervention. If a change in treatment by the village health worker, or the introduction of a special form of health education, is the intervention it may be imposs-ible to randomize within the village. In these situation