In summary, it is impossible to reconcile the contradictory results of the studies of the association between previous blood transfusion and subsequent development of NHL (Figs 1 and 2). Most of the studies were of high quality, and the authors discussed the possible sources of bias that might have generated a spurious association or concealed a true relationship. Although recall bias, detection bias, and allogeneic blood transfusions given because of the onset of NHL could be responsible for the reported positive associations, these explanations were considered by the investigators, and they seem unlikely. Multiple comparisons may account for the association between transfusion history and NHL in individual studies, but they are less likely to explain the association noted in all the positive studies considered together. Random misclassification and Berkson's bias might explain some of the null results, but they are unlikely to account for disagreements of the magnitude shown in Figures 1 and 2. Alexander discussed the superior quality of the design of the case-control study of Adami et al but was unable to explain the disagreements among the reported studies (Figs 1 and 2). He concluded that there is no proof that allogeneic blood transfusion does not increase the risk of NHL; one can, however, be reassured that the evidence so far points to--at worst--a doubling of risk, and--at best--no increase in risk after a previous transfusion. Along these lines, Figure 1 shows the results of the 3 cohort studies, which were remarkably consistent in reporting a 2-fold increase in the risk of developing NHL after a blood transfusion; and Figure 2 shows the results of the 6 case-control studies, which usually observed no increased risk. If allogeneic blood transfusion does have an immunosuppressive effect, this effect is probably transient and weak, compared with the severity of the immunosuppression encountered in the posttransplantation situation. Immune impairment may be the common determinant for the increased risk of NHL observed in transplanted and transfused patients, and--if this were the case--the difference in the duration and intensity of the immunosuppressive state would be logically congruent with the observed patterns of lymphoma development: that is, the risk of NHL is increased 20- to 120-fold in the transplanted patients, as compared with only 2-fold in the previously transfused patients. The association between allogeneic blood transfusion and subsequent development of NHL is biologically plausible, whether the mechanism is transfusion-associated immunosuppression, transmission of oncogenic viruses, or viral activation in a setting of transfusion-associated immunosuppression. Also, if allogeneic blood transfusion is a risk factor for the subsequent development of NHL, the increased use of allogeneic blood transfusion since the 1950s might account, at least in part, for the increase in the incidence of NHL over the last half of this century. Blood transfusions are commonly used worldwide, and--based on at least some studies--they show a weak association with NHL (i.e., at worst, a doubling of risk; Fig 1). Common exposures that have a weak association with NHL are more likely to account for the current epidemic of NHL in the elderly, compared with rare exposures that increase the risk of NHL by manyfold. However, no evidence regarding a causal relationship between history of an allogeneic blood transfusion and the subsequent development of NHL has been presented. The available studies are observational, and they cannot determine whether any increase in risk, observed in association with allogeneic blood transfusion, is due to the transfusion itself or to other factors occurring in association with the transfusion. In conclusion, allogeneic blood transfusion from healthy donors may be associated with a small increase in the risk of development of NHL after the transfusion. This hypothesis is biologically plausible and is