Letters

Patients who are eligible but not randomised should be included as additional comparative arm in study

EDITORPeto and Baigent state that we need to find ways of making trials much simpler and larger.¹ We agree and have a suggestion based on 20 years' experience.²

Everyone is familiar with inclusion and exclusion criteria for randomised controlled trials. With rare exceptions, however, those patients who are eligible but not randomised are a forgotten part of the population; this presents a potentially large problem for all trials. The principal investigators of randomised controlled trials often find that the number of patients being entered into a trial by a particular contributor is not related to the size of the population served by that doctor. There are many reasons for this: purported logistic problems of data entry, inconvenience for patients, patients' unwillingness to accept randomisation, non-specific unsuitability of patients for the trial, etc. The doctor's inability or unwillingness to offer the trial to all eligible patients provides the circumstances in which clinical selection bias is likely to have occurred.

We recommend that in all trials there should be an absolute requirement that the "eligible but not randomised" group of patients should be included in the documentation process as an additional comparative arm in the study. To investigate the possibility that clinical selection has biased the trial results, this group should be compared with the true control group. If the outcome results are similar then we may predict that the result of the trial will apply to the general population. By contrast, if the outcome results are different then selection bias has clearly occurred, such that the applicability of the trial results to an unselected population must be questioned.

Furthermore, routine documentation of all patients being cared for in departments and units undertaking randomised trials would confer some additional benefit. For example, the larger number of patients being documented and followed up would allow variance between treatments and between outcomes achieved by different doctors to be recognised. The study of such variances will complement the results of randomised controlled trials as we search for reliable new knowledge.

Insisting that all eligible patients be documented would have several consequences: the size of the eligible but not randomised group would be minimised; the size of the trial itself would be maximised; and the results of the trial would be more readily applicable to the population at large. This strategy would help in the recruitment of the necessary numbers of patients for meaningful results and reduce the time taken to complete the trial. Achieving both these objectives would enhance the value of all trials.

L Peter Fielding, Director, surgical services. 
York Health Surgical Services, 1001 South George Street, York, PA 17405, USA PFIELDIN{at}yorkhospital.edu

Roger Grace, Director. 
Division of Clinical Science, University of Wolverhampton, Wolverhampton WV1 1SB

Rosemary Hittinger, Senior clinical audit coordinator. 
Department of Clinical Audit, Acrow Building, St Mary's Hospital, London W2 1NY

1. Peto R, Baigent C. Trials: the next 50 years. BMJ 1998; 317: 1170-1171[Free Full Text]. (31 October.)
2. Fielding LP, Stewart-Brown S, Dudley HAF. Surgeon-related variables and the clinical trial. Lancet 1978; ii: 778-779.