Independent review says the patient is ineligible

Many trials will incorporate central review of eligibility, for example pathology or radiology, where a small group of experts will review all the patients in a trial and confirm, or not, their eligibility according to the protocol. Whether or not this has the potential to introduce bias depends on whether the reviewers have information on the patients' treatment or outcome which could conceivably influence their diagnosis. Of course anyone intent on bias could, in the knowledge of treatment allocation, declare ineligible all those patients in one treatment group with the features indicating poor prognosis. A more subtle bias may creep in if the reviewer has knowledge of the patients' outcomes. For example, a neuropathologist reviewing slides for brain tumour patients entered into a trial of high-grade glioma may, on the basis of the limited material available to him, be unsure whether some patients have a grade 2 (low grade) or grade 3 (high grade) tumour. If the review was being done retrospectively, and he was aware of the patients' survival times, he may be steered towards a diagnosis of grade 2 (ineligible) for those with longer survival and grade 3 (eligible) for those with shorter survival, since this would be in keeping with the usual prognosis of these patients. If in fact the reason for the longer survival for some patients was that they had been allocated a particularly effective treatment, their exclusion would bias the results against this treatment. These issues are avoided if central reviews are conducted 'blind' to treatment allocation and outcome, the latter being helped by conducting the review prospectively. Where review is blind, the question is whether it is appropriate to exclude a patient on the basis of an expert opinion, when that opinion would not routinely be available to patients outside the trial. This issue is discussed further in Section 4.2.3.

The difficulty with all but the last of the situations outlined above is that the reason for exclusion is very likely to be related to the treatment that the patient has been allocated, leading to the potential for systematic differences between the experimental and control groups being analysed. Unless we can be absolutely sure that the reasons for exclusion are completely independent of the allocated treatment, then the best, least biased and most conservative approach is to include all patients in the analysis. This is called the 'intention-to-treat' or 'ITT' analysis (see Box 9.2) defined as including all randomized patients and analyzing them according to the treatment to which they were allocated irrespective of the treatment that they actually received. In contrast, examples of when exclusion of patients might be justified are when an independent review has been done of pathology, blind of the treatment allocated and also in 'equivalence' trials, and this is discussed in Section 9.4.2.

0 0

Post a comment