Red for danger: blood transfusion and colorectal cancer.

The ability to differentiate the value of a clear set of observations from the inferences which may be drawn from them is the hallmark of a physician with "good clinical judgment." This type of discernment is also needed for retrospective observational clinical studies. Prospective controlled trials should investigate a hypothesis which has some a priori reason to support it, but those concerned with retrospective analyses have to be satisfied with generating new ideas to be tested, A good recent example of this process comes from several studies which have shown that patients with colorectal cancer survive longer if they are not given blood tranfusions at the time of their surgery.` The research workers have not only suggested that this finding may be causally related but also considered the possibility that their observation might be an epiphenom-enon: it might be that in each histopathological stage those patients needing blood transfusions had an intrinsically worse prognosis than those patients in the same stage who did not need them. Possibly a cumulative effect in the population as a whole has simply identified those patients with an inherently poorer outlook. The size of the difference in prognosis suggests, however, that this observation is not just a statistical statement of minor clinical importance but may be one of the crucial factors determining outcome, with an up to fourfold effect on tumour recurrence between the two groups' and a large influence on survival.2 Furthermore, their conclusion that perioperative blood transfusion has an independent influence on prognosis is supported by the statistical methods used-namely, the log rank and Cox proportional hazard regression analyses, which allow the more usual single variate analysis to be followed by multi-variate adjustments made for all significant covariance simultaneously.' Yet if it is assumed that the differences observed were directly related to blood transfusion the biological basis for this outcome may be difficult to explain. A simple analogy to the beneficial effects of blood transfusion in association with kidney transplantation is attractive. That would suggest that the transfused blood (probably the white blood cells) results in immunosuppression, releasing the tumour to metastasise and grow. Several different types of immunological events have been ©t BRITISH MEDICAL JOURNAL 1985. All reproduction rights reserved. described after blood transfusion in patients given renal transplants. One possibility is that the beneficial effects of blood transfusion on the survival of kidney grafts may be attributed to clonal deletion and …